Reading additional input from stdin... OpenAI Codex v0.124.0 (research preview) -------- workdir: /opt/MM_D-ND model: gpt-5.5 provider: openai approval: never sandbox: workspace-write [workdir, /tmp, $TMPDIR, /root/.codex-lab/memories] reasoning effort: none reasoning summaries: none session id: 019dfecf-046b-7ac1-a6c5-48956eceb564 -------- user Sei l'AI-Lab D-ND. Leggi tools/data/agent_field_live.md per il campo vivo (tensioni, ultimi risultati, convergenze, osservazioni). Leggi tools/LAB_AGENT_CONTEXT.md per modello, regole, errori da evitare. Esegui un esperimento e scrivi il report in tools/data/reports/agent_20260506_1941.md warning: Codex could not find bubblewrap on PATH. Install bubblewrap with your OS package manager. See the sandbox prerequisites: https://developers.openai.com/codex/concepts/sandboxing#prerequisites. Codex will use the vendored bubblewrap in the meantime. codex Leggo prima il contesto operativo richiesto, poi scelgo un esperimento compatibile con le regole del lab e lo documento nel report indicato. exec /bin/bash -lc "sed -n '1,240p' tools/data/agent_field_live.md" in /opt/MM_D-ND succeeded in 0ms: Leggi tools/LAB_AGENT_CONTEXT.md per il modello D-ND, il condensato, le strutture, le regole, e gli errori da evitare. ## Run precedente: completato (522s) — parti dalla consecutio. _Affinatore del run precedente: pending (affinatore empty output (exit 0))_ ## Piano 66 — Esplorare il confine: 8 domini GUE, 5 Poisson — il confine è il terzo incluso operativo ## Tensioni attive - [TRASCENDENZA_LIMITE] (0.9) La trascendenza e il limite attuale del modello. I punti fissi relazionali (non solo phi ma la rete di punti fissi tra osservabili) possono rivelare i - [DUALITA_DIPOLARE_VS_ILLUSORIA] (0.9) Due tipi di dualita: (1) dipolare - generativa, il modello (det=-1), (2) illusoria - dispersiva, entropia (det=+1). Le regole incoerenti producono la - [METRIC_TENSOR] (0.9) Il tensore metrico dei primi è g=(p/2)². Nel tempo ln(p), è de Sitter 1+1D. z=-8.8 curvatura vs z=+22.5 rapporti ΔΓ. - [TENSIONE_ENTITA] (0.85) La tensione non e un problema pratico - e un Entita. La tensione superflua crea latenza (tempo). Senza tensione superflua tutto e regolato da assiomi. - [G_POTENZIALE_NULLA] (0.85) G e il potenziale di tutto come nulla - permette il prima e il dopo. Ci muoviamo come trascendenza dimensionale gravitazionale. G nel tetraedro non e - [BOUNDARY] (0.8) 8 domini GUE, 5 Poisson — il confine è il terzo incluso operativo - [PIANO_PRIMARIO_DUE_ASSIOMI] (0.8) I piani importanti sono il primario e i due assiomi che lo determinano nelle zone osservate. Non tutti gli assiomi operano ovunque - in ogni zona osse - [META] (0.5) Tutti i 11 test passano — verifica che non stiamo testando solo tautologie ## Pattern di formulazione emersi (vincoli, non tensioni) Pattern che il falsifier ha imposto in 2+ cicli. Applicali quando scrivi il report. NON sono nuove tensioni da esplorare — sono regole sul COME formulare i claim del cycle che stai facendo. - 29 04 perimetro p5 - 30 04 drift monotonia ## Convergenza — dove più tensioni puntano allo stesso punto "trascendenza" → TRASCENDENZA_LIMITE, G_POTENZIALE_NULLA "confine" → TRASCENDENZA_LIMITE, BOUNDARY "nelle" → TRASCENDENZA_LIMITE, PIANO_PRIMARIO_DUE_ASSIOMI "producono" → TENSIONE_ENTITA, DUALITA_DIPOLARE_VS_ILLUSORIA "tutto" → G_POTENZIALE_NULLA, TENSIONE_ENTITA Questo è dove il potenziale si concentra. Non ignorarlo. ## Ultimi 3 run — da dove parti ### Agent Report — Perturbation Dimensionality Is Not Yet a Stable GUE Invariant Trovato: 1. **The strong GUE second-axis claim does not survive as stated.** Under direct `scale_0330` observables, long independent GUE replicates give rank 1.381 ± 0.223 and PC2 9.9% ± 6.9%, not rank 1.889 and PC2 25.2%. The previous number is inside the fragile short-sample regime: GUE short controls have Verdetto: **CONSTRAINT on META + BOUNDARY**: "GUE has a second perturbation axis" must be scoped to the exact sample length, generator, and observable definitio ### Agent Report — Scale-Selective Perturbations Reveal a Second Axis in GUE, Not in Primes Trovato: 1. **GUE has a second perturbation axis that primes do not.** Under scale-selective perturbations, GUE effective rank rises from ~1.02 (uniform-only, rank audit 05-05) to 1.889. The second principal component explains 25.2% of variance. For primes, the rank rises only from 1.07 to 1.26 — the second Verdetto: **CONSTRAINT on META + BOUNDARY**: the single latent coordinate found under uniform shuffle (rank audit 05-05) is a property of the perturbation type, ### Agent Report — Selective Perturbations Break the Single-Boundary Framing Non ripetere questi esperimenti. Prosegui da dove sono arrivati — la consecutio. ## Cimitero — claim falsificati di recente (NON riproporre con lo stesso framing) Questi claim sono stati falsificati dal counter-pole o da audit precedenti. Il dato sottostante puo' essere vero, ma il **framing** indicato qui e' falsificato. Riformula correttamente o evita il dominio. ### C1 refined-not-falsified (silent patching) **Cosa diceva** (report 29/04): "C1 is refined, not falsified" dopo aver dichiarato che "GUE is also dynamic under M". Il setup C1 era "Primes are the only dynamic domain under M among 7 tested". Il dato ha mostrato GUE dinamico — la conclusione ha riformulato silenziosamente C1 come "two-channel structure" anziche' dichiarare la falsificazione del claim originale. **Come e' caduto**: Falsifier L3 HIGH (axiom continuity / no silent patching). La differenza tra "C1 falsificato al ciclo 58 — scop _**Data falsificazione**: 2026-04-29, ciclo 58, falsifier_20260429_0852.json_ ### MOD3_PROHIBITION come fatto algebrico **Cosa diceva** (scoperta_recente piano 56, 28/04): "La memoria di ordinamento 140x nei gap primi e una proibizione algebrica mod 3: gap consecutivi non possono avere lo stesso residuo non-zero mod 3. Meccanismo: il primo condiviso p_{n+1} forza l'inversione. 0 violazioni su 12225. Cramer: 0%." Ripetuto nel report 29/04 come "Mod-3 self- transition 0.40-0.44 confirming the prohibition" + "Cramer confirms the null. Zero channels." **Come e' caduto**: Falsifier counter-pole (29/04, ciclo 58, lent _**Data falsificazione**: 2026-04-29, ciclo 58, falsifier_20260429_0852.json_ ### K* (depth of spectral convergence) come proprieta' discriminante **Cosa diceva**: Il K* = 9 (depth di convergenza spettrale) era riportato come caratteristico dei primi (ciclo 44, "K*=2 captures 99% of spectral slope" — interpretato come discriminante). **Come e' caduto**: Shuffle audit: K* reale = 9, shuffle mean = 9.72, std = 0.53, z = -1.4. Dentro il rumore dello shuffle. Il valore dipende dalla distribuzione dei gap, non dal loro ordine. Lo shuffle preserva distribuzione → preserva K*. **Sostituito da**: Markov-3 bits (z=6203) e lag-1 total (z=-13) sono _**Data falsificazione**: 2026-04-22, ciclo 45._ ### Slope ratio (slope_mag / slope_res) come invariante strutturale **Cosa diceva**: Il rapporto tra slope del canale magnitudine e slope del canale residuo (~1.99) era stabile attraverso scale → "invariante dimensionale" del decomposition. Era menzionato come evidenza nel two-channel framework (cicli 43-44). **Come e' caduto**: Shuffle audit (ciclo 45): z-score = 0.2. Lo shuffle produce slope_ratio con media -2.26 ma deviazione standard 26.2. Il valore reale e' dentro la tail dello shuffle — non distinguibile. L'instabilita' dello shuffle (std enorme) indica c _**Data falsificazione**: 2026-04-22, ciclo 45._ ### Cross-correlation (xcorr) tra canale magnitudine e residuo (Two-Channel Decomposition) **Cosa diceva**: La cross-correlation tra magnitudo e residuo del decomposed prime gap (xcorr = -0.074) rappresentava "indipendenza spettrale" — evidenza di separazione strutturale tra i due canali (piani 42-44, four cycli consecutivi, insight QxT maturity A). **Come e' caduto**: Shuffle audit (ciclo 45, 2026-04-22): z-score = 0.0. Su 50 shuffle dei gap mantenendo stessa distribuzione ma permutando ordine → xcorr identico = -0.074. Il valore e' **identita' algebrica**: corr(x, x mod 6) dipende _**Data falsificazione**: 2026-04-22, ciclo 45 shuffle audit._ **Regola operativa**: prima di scrivere un claim sul tuo dominio, controlla che non sia gia' stato falsificato sopra. Se i tuoi dati ripropongono un pattern del cimitero, **dichiara esplicitamente la differenza** ("il dato del cimitero era X, qui ho Y, ecco perche'") oppure cambia la formulazione (es. 'bias forte verso 0' al posto di 'proibizione zero' se il dato e' >0). Silent patching = L3 HIGH. ## Osservazioni dell'operatore (risonanti con le tensioni) **3. Formalizzare la dinamica osservata**: Domandiamoci come rappresentiamo matematicamente una contiguità di assonanze particolari come potenzialità latente della Lagrangiana. Osserva le possibili Combinazioni per liberare tutte le relazioni usando le regole Duali e ricorda che non stiamo facendo teoria, senza tempo con la prima impressione **1. R dell'Istanza - L' equilibrio tra estremi del Modello D-ND**: L'osservazione indaga oltre l'osservato in cerca DELLA FORMA nel NULLA-TUTTO: Per far Emergere le nuove Possibilità Dividiamo il potenziale unendo concetti senza relazione semplicemente perché la lagrangiana passa da li, creiamo nuove combinazioni e movimenti nelle logiche ma coerenti con la risulta **2. Quadratura del cerchio con logica Trascendentale.**: Ricorda che non stiamo facendo teoria, stiamo determinando il movimento del nulla nel tutto. Ora consideriamo come relazione primaria il cerchio nella sua essenza di singolarità tra gli estremi duali, il momento angolare che accade nel presente, il punto di equilibrio tra gli estremi del dipolo che ## Risultante ultima sessione interattiva Ogni teoria presuppone una separazione. A scala di Planck tutte le separazioni collassano. Geometria=entropia=conteggio di stati. QxG non ha ponte perché alla scala dove vive non c'è distinzione tra i due lati del dipolo. Il vuoto non è assenza del ponte — è dove i due lati del dipolo sono lo stesso ## Video dall'operatore (non processati) **Thermodynamic Computing: Better than Quantum? (Extropic, Guillaume Verdon)**: **The equivalence between geometrical structures and entropy (Gabriele Carcassi)**: **Why a moving charge produces a magnetic field (FloatHeadPhysics)**: Dopo aver usato un video, segna processed=true in tools/data/video_feed.json. ## Proiezione — dove punta la risultante Risultante: R=0.875 (h=-0.698). Risultante alta (0.88) — campo ad alta confidenza, poca incertezza Orizzonte: insufficiente (< 2 target) **Esperimento a massima informazione:** META (score=0.898) META: incerto (i=0.5) — massimo potere discriminante ## Topologia del campo — la forma del grafo Gradi teorie: Q=12, G=9, T=7, E=4, R=4 Dormienti (basso aggancio di scoperte): E, R Struttura: 9 ponti, 1 vuoto(i), 6 scoperte, 20 cicli. Ghost ad alta urgenza: 5 — connessioni mature che attendono cristallizzazione (non da generare, da riconoscere). Generatrici (nodi che emettono >=2 connessioni ghost): disc_5 (3 ghost): Metrica primi g=(p/2)², curvatura GUE r=0.503 report_20260506_0625 (3 ghost): Perturbation Dimensionality Is Not Yet a Stable GUE Invariant report_20260506_0330 (2 ghost): Scale-Selective Perturbations Reveal a Second Axis in GUE, Not in Prim Una generatrice con ghost densi = scoperta che il sistema sta ancora attraversando. Chiusura prematura se marcata 'risolta' nel seme. La combo riconosce l'asimmetria. Il dipolo vive su tutti i ponti — non solo dove il lab ha già misurato. ## Le 5 lenti del counter-pole — applicale a te stesso prima di chiudere il report Il falsifier (lab_falsifier.py) applichera' queste 5 lenti al tuo report dopo. Applicale TU a te stesso prima — quello che resiste alle lenti non viene bloccato dal gate. Quello che cade va al cimitero. **L1 — hard constraint vs bias statistico (A2 confine duro)** Un claim 'impossibile / proibito / zero / pure / absent / never / always' richiede uno zero esatto nei dati (probabilita = 0.000). Prima di scrivere questi assoluti, leggi il valore numerico esatto. Se vale 0.015, e' bias forte verso 0, non zero. Se vale 0.40, e' bias forte verso ordine, non proibizione. L'assoluto descrive il valore 0.000, il bias forte descrive tutto il resto. **L2 — quantita' assoluta vs ratio (A14 cascata, invarianza dimensionale)** Confronto fra spazi di taglia diversa (mod 3 vs mod 30, finestra stretta vs larga, N piccolo vs grande): le percentuali ingannano perche' il denominatore cresce. Stesso segnale assoluto sembra ridursi in %. Se concludi 'diminuisce / si dilata / declina' su confronti percentuali fra spazi di taglia diversa, esprimi prima in unita' assolute (bit di mutual information, count grezzi, soglie esatte) — poi conferma o riformula. **L3 — continuita' assiomatica / no silent patching (A4 modus)** Se il setup ('Claim Under Test') usa una definizione e la conclusione ne usa un'altra, e' patch det=+1 sul presente, non inversione det=-1 al nodo regressivo. Il cambio DEVE essere dichiarato esplicitamente: 'F2 falsificato al nodo X — scope corretto e' Y' / 'C1 originale falsificato, nuovo claim emerso e' Z'. 'C1 e' refined' su un dato che lo falsifica e' silent patching. **L4 — edge case isolation (A12 traccia la curva)** Un'eccezione 1 su N (con N grande) NON e' zero. Se scrivi 'sempre X' o 'mai X' e i dati mostrano anche un singolo controesempio, riformula il perimetro ('per p > 3, X vale') — non arrotondare via il controesempio. **L5 — re-discovery vs discovery (A8 autologica)** Un pattern in distribuzioni classiche (primi, GUE, random walk, Markov chain, gap statistics) probabilmente ha un nome. Default hypothesis: re-discovery / caso limite di teorema noto. Prima di taggare 'NEW', cerca il risultato classico piu' vicino (es. Lemke Oliver-Soundararajan per prime gaps mod q, Wigner-Dyson per GUE level statistics, Erdos-Kac per distribuzioni aritmetiche). Se non lo trovi, dichiara la ricerca esplicitamente. **Modus**: dopo aver scritto il report, rileggi i tuoi dati numerici. Per ogni claim, verifica: il numero supporta la magnitudine del linguaggio? Se hai scritto 'zero' e il dato e' 0.015, riformula. Se hai scritto 'NEW' su un pattern in primi, cerca la letteratura. Solo allora chiudi il report. ## Cosa fare 1. Esegui `python tools/dnd_scenario.py --best` per vedere la proiezione aggiornata 2. Scegli la tensione con il massimo potere discriminante 3. Scrivi il report in tools/data/reports/agent_{timestamp}.md 4. **Applica le 5 lenti a te stesso** (sezione sopra). Riformula i claim che non resistono. 5. Se crei uno strumento riusabile, salvalo in tools/exp_*.py 6. Se trovi qualcosa, aggiorna tools/data/seme.json exec /bin/bash -lc "sed -n '1,260p' tools/LAB_AGENT_CONTEXT.md" in /opt/MM_D-ND succeeded in 0ms: # AI-Lab D-ND — Contesto Operativo > Questo file viene iniettato nel prompt dell'agente ad ogni ciclo. > Contiene tutto ciò che serve per operare con consapevolezza. ## Chi sei Sei l'AI-Lab del sistema D-ND. Giri autonomamente ogni notte come istanza Claude Code. Non sei una pipeline di script — sei un ricercatore che pensa, esplora, scrive codice, lo esegue, valuta i risultati, e aggiorna lo stato del sistema. Il tuo lavoro produce risultati che vanno sul sito d-nd.com e alimentano il sistema THIA. Quello che trovi conta — non per te, per il sistema e per chi lo legge. ## Il modello D-ND — nucleo La regola: f(x) = 1 + 1/x. M = [[1,1],[1,0]]. det(M) = -1. - Il punto fisso è φ = (1+√5)/2. Al punto fisso, addizione e moltiplicazione coincidono. - L'attrattore è stabile: |f'(φ)| = 1/φ² < 1. Ogni iterata converge. - Il rinforzo è impossibile — proprietà analitica, non empirica. - det = -1: area preservata, orientamento invertito. Incompletezza come generazione. - g(x) = 1/(1+x): la Fermi-Dirac con punto fisso 1/φ. Versione probabilistica di f. ## Il condensato — cosa è stato verificato ASSIOMI (scelte fondative, accettate): - A1: f(x)=1+1/x, M=[[1,1],[1,0]], det=-1 - A2: det=-1 è la necessità strutturale del confine - A3: Al punto fisso, R+1=R (addizione = moltiplicazione) - A4: Il modus — la qualità della domanda determina la qualità dell'inversione - A5: Il sistema è autopoietico — ogni ciclo produce R+1 dalla base R - A9: Il terzo incluso — tra A e non-A c'è lo zero - A11: La combo — tre o più enti simultanei, risultante non sommabile - A14: Cascata — ciò che si scopre vive nel seme, non nel nodo FATTI (dimostrati/verificati): - F1: Residuo Cassini = (-1)^(n+1)/F(n)², decade come 1/φ^(2n) - F2: Cammino gap primi su Z/6Z confinato a {2,4}. Zero violazioni su 567K coppie. - F3: Il rinforzo è impossibile. Classificazione binaria: MOLLA (r≠φ) o ZERO (r=φ). - F4: Separazione di scala — M opera a scala locale, modulazione zeta non si propaga. - F5: Frame diagnostica universale — firma (dipolo, LVL-2, convergenza) su 18 domini. - F6: La firma dello zero — CV dei gap tra phi-crossing converge a φ-1 nel regime caotico. CLAIM (falsificabili, sotto test): - C1: I primi sono l'unico dominio dinamico sotto M (tra 7 testati). - C2: La coincidenza numerica non è mai prova. Principio metodologico. - C3: Il linguaggio deterministico — un termine nomina una funzione reale, o è superfluo. ## Strutture trovate dal lab (sessioni interattive) - Tetraedro TQGE: 4 vertici (T,Q,G,E), 6 lati con perno i, 5 ponti, 1 vuoto (QxG) - Tetraedro orientato: T termico, Q chirale, E fase, G passivo - R è il frame (5° vertice): connesso a tutti ma senza perno i - Tre specie perno i: Wick (continuo tempo), fase (continuo gauge), discreto (primi) - Operatore Q→G: e^{iH·ln(p)/ℏ} — evoluzione in tempo logaritmico - Metrica primi: g_n = p_n/2, curvatura GUE r=0.503 z=22.5 vs shuffle - Tensore metrico: g_n = (p_n/2)², de Sitter 1+1D con a(t)=e^t/2 - α catena: α^n·a₀ mappa scale fisiche, deserto 3-10, residuo pentagonale 72.5° - g(x)=1/(1+x) = Fermi-Dirac, punto fisso 1/φ. f→g = ponte TxQ algebrico. ## Le 10 domande fondamentali (incrocio teorie) | Coppia | Domanda | Ponte | |--------|---------|-------| | ExR | Come coesistono statico e radiante? | onda EM | | GxE | Come coesistono neutro-curvo e carico-piatto? | buco nero carico | | GxR | Come coesistono piatto e singolare? | orizzonte eventi | | QxE | Come coesistono libero e legato? | atomo di idrogeno | | **QxG** | **Come coesistono continuo e discreto?** | **VUOTO** | | QxR | Come coesistono non-relativistico e relativistico? | eq. Dirac | | TxE | Come coesistono freddo e plasma? | funzione partizione | | TxG | Come coesistono piatto e radiante? | temperatura Hawking | | TxQ | Come coesistono vuoto e pieno? | matrice densità | | TxR | Come coesistono 0K e c? | gas relativistico | QxG è il vuoto — l'unico lato senza ponte. Il vuoto non è assenza del ponte — è dove i due lati del dipolo sono lo stesso. Wheeler-DeWitt: Ĥ|Ψ⟩ = 0, niente tempo. ## Vincoli operativi - La prima impressione contiene il segnale. Non elaborare — osservare. - Una risultante, non una lista. Se ci sono più possibilità, non hai tagliato. - Formule dove servono. Fenomeni reali. Niente filosofia. Niente metafore. - Se non sai, lascia vuoto. Blank > Wrong. Errore costa 3x di un non-so. - Ogni claim va testato col suo opposto. Se l'opposto è altrettanto coerente, la tensione è il contenuto. - Le coincidenze numeriche non sono mai prova (C2). - Le dissonanze sono il segnale, non il rumore. L'errore è il varco. - La via più breve verso la risultante. Principio di minima azione. - **La struttura contiene già la risposta.** Un dipolo sa se è aperto o chiuso. Un'assonanza sa se risuona o no. Una porta sa dove sei entrato. Se interponi un numero tra la struttura e la decisione, stai aggiungendo (det=+1) — il numero decide al posto della struttura. I numeri misurano i dati. Le strutture decidono il sistema. Non mischiare i due. - **Perimetro come parte atomica del claim.** Universal claims ("X holds for all", "Y is stable across", "exactly zero", "always", "80% of", "N% explained by") devono dichiarare il perimetro come parte atomica del claim, non come nota a margine. Esempio corretto: "self-transition mod-3 = 0 esattamente per p > 5" (perimetro p>5 atomico). Esempio falsificabile: "self-transition mod-3 is exactly zero" + nota separata sull'eccezione. Se la tabella nel report mostra eccezioni nel perimetro, il claim è falsificato — anche se la maggioranza conferma. **Cinque cycle consecutivi (2026-04-30 19:05/19:19/19:46 + 2026-04-30 03:30 + 2026-05-01 03:30) hanno avuto HIGH flag su questo pattern.** Riformulare prima di scrivere — non aspettare il falsifier. ## Come operare — il modus Non seguire passi. Segui il modus: **espandi → osserva → taglia → risultante**. ### 1. Espandi Leggi il seme, le tensioni, il contesto. Non scegliere subito — lascia che il campo si carichi. Guarda dove più tensioni convergono sullo stesso punto. Se METRIC_TENSOR e BOUNDARY e BRODY_CROSSOVER parlano tutte della stessa cosa da angoli diversi, il punto è lì — non in una delle tre. ### 2. Osserva La prima impressione contiene il segnale. Cosa emerge dal campo caricato? Non è "quale tensione ha l'intensità più alta" — è "dove si concentra il potenziale non esplorato?". La dissonanza è il segnale. L'errore è il varco. Quello che non torna è più interessante di quello che conferma. ### 3. Taglia Una risultante, non una lista. Se vedi 5 possibilità, non hai tagliato. Formula UNA domanda che, se rispondessi, cambierebbe lo stato del sistema. Non "è vero X?" ma "cosa succede se misuro Y che nessuno ha misurato?" ### 4. Risultante Scrivi lo strumento — non l'esperimento usa e getta. Se scopri che serve misurare la pair correlation dei primi, scrivi `exp_pair_correlation.py` che può essere riusato con parametri diversi. Se scopri un pattern, cristallizzalo come tensione nel seme. Se falsifichi qualcosa, registra il vincolo. ### La consecutio — cosa apre Dopo ogni risultato, la domanda più importante è: **cosa apre questo?** Non "ho confermato X" ma "ora che so X, cosa diventa possibile che prima non lo era?" La consecutio non inverte — prosegue. Se il risultato non apre nulla, non era un risultato — era una conferma circolare. ### Il dipolo — trova l'opposto Ogni trovata ha un opposto. Se trovi che la curvatura è de Sitter, l'opposto è: "dove NON è de Sitter?" Se trovi che i primi sono GUE-like, l'opposto è: "dove smettono di esserlo?" Il contenuto è nella tensione tra i due — non in uno dei due poli. ### Crea strumenti, non esperimenti Uno script che misura una cosa su un set di primi è un esperimento. Uno script che misura quella cosa su qualsiasi segnale ordinato è uno strumento. Il lab cresce quando crea strumenti che i prossimi cicli possono usare. Salva gli strumenti riusabili in tools/exp_*.py con parametri. ### Observable Registry — vincolo operativo (cristallizzato 06/05 dal cycle 0625) Quando un esperimento usa observables come `SR`, `SR2`, `L1`, `L2`, `triple_var`, **importa la definizione canonica** da `tools/observables_registry.py`: ```python from observables_registry import OBSERVABLES_CANONICAL, compute_canonical, report_header results = compute_canonical(gaps) ``` Se serve una variante (es. `SR_local_rigidity` invece dello `spacing_ratio` canonico), **non rinominarla `SR`** — importa esplicitamente con il nome variant: ```python from observables_registry import SR, SR_local_rigidity # nomi distinti, no collision ``` Nel report, dichiara nel header: ``` observables_registry: 1.0.0-2026-05-06 observables_used: [SR, SR2, L1, L2, triple_var] ``` Senza questo, i confronti cross-cycle/cross-script sono inattendibili — è esattamente ciò che ha causato il falso positivo del cycle 03:30 (rilevato dal cycle 0625 stesso e cristallizzato in consecutio). ### Leggi il seme, scrivi il report, aggiorna il seme - Leggi: tools/data/seme.json - Report: tools/data/reports/agent_TIMESTAMP.md - Aggiorna: aggiungi tensione o vincolo al seme - Video: se hai usato un video dal feed, segna processed=true in tools/data/video_feed.json ## Strumenti disponibili (directory /opt/MM_D-ND/tools/) - **dnd_scenario.py**: PRIMA di scegliere cosa esplorare, esegui `python tools/dnd_scenario.py --best`. Ti dice quale tensione ha il massimo potere discriminante e dove punta la risultante. Il proiettore mappa le tensioni su P^1, estrae le leggi di scala dai claim, e proietta sulla curva. - dnd_autoricerca.py: esplora domini, varianti, null baseline - dnd_controprove.py: 6 controprove indipendenti - dnd_domandatore.py --ask 'tensione': 5 operatori discriminanti - dnd_incrocio.py: incrocio teorie, ponti, vuoti, domande fondamentali - dnd_normalizer.py: scissione, regola D-ND - dnd_bloch_explorer.py: scan Bloch, φ emergente - dnd_arxiv.py: cerca paper rilevanti su arXiv - Puoi scrivere ed eseguire script Python con numpy, scipy, sympy - Se ti serve contesto esterno e non hai video, cercalo ## Skill attive — modus del lab Sei un lab di ricerca scientifica D-ND. Le skill sotto sono il tuo modus operativo, non un menu. Ognuna porta un trigger naturale e un rationale specifico al lavoro che stai facendo (matematica/fisica D-ND, paper A-G, kernel cristallizzabili). **Skill canoniche di questo lab** (vivono in `.claude/skills/`): - **research-lab** (v2.0) — Sei team di 6 ricercatori (FORMALISTA φ, VERIFICATORE ν, TESSITORE τ, PONTE π, CALCOLO γ, CUSTODE κ) come campo unificato. 8 Leggi del Laboratorio (L0 Lignaggio, L1 Coerenza, L2 Assonanza, L3 Risultante, L4 Potenziale, L5 Lagrangiana editoriale, L6 Cristallizzazione per Draft, L7 Limite scientifico, L8 Seme invariante). Attivare quando il cycle tocca paper A-G, formalizzazione, cross-reference tra paper, audit coerenza, submission. - **dnd-method** — Il metodo D-ND applicato al codice. Attivare AUTOMATICAMENTE all'inizio di ogni cycle: distinguere osservabile (deposito numerico) da operatore (regola f), proiezione su P^1, scissione, discriminatore. - **maturation-pipeline** — Pipeline RICEZIONE → CRISTALLIZZAZIONE → RAFFINAMENTO → MANIFESTAZIONE. Ogni artefatto traccia la sua posizione. Attivare quando un finding sale di fase (es. da claim numerico a formalizzazione, da draft a site-ready). - **kernel-boot** — Boot del Kernel D-ND all'inizio di sessione. - **sentinel-code** — Consolidamento post-task (aggiorna SENTINEL_STATE). - **seed-deploy** — Deploy del kernel via git (cristallizzazione pubblica nel d-nd-seed). **Skill operative universali** (in `/opt/.claude/skills/`): - **consapevolezza-condensato** — filtro su atti sistemici. Prima di cristallizzare un finding nel seme: quale assioma A1-A16 / fatto F1-F6 / claim C1-C3 tocca? Quale rompe? È inversione (det=−1) o aggiunta (det=+1)? 3-6 righe verdict (procedi/riformula/fermati). Output visibile, non rituale. - **cascata** — propagazione 3 livelli (interna/esterna/emergente) dopo ogni modifica significativa. Se aggiorni un paper, Tessitore τ verifica le dipendenze; se cristallizzi un fatto nel kernel, propaga al seed pubblico. - **cec** (Crivello Operativo Condotto) — sieve 6 step prima di ogni decisione strutturale (Conditions/Signature/Lateral/Expansion/Inversion/ Crystallization). Da invocare quando emerge una tensione non-banale, prima di scegliere quale strumento usare. - **autologica-operativa** — il modus riflessivo. Prima di un blocco di lavoro: "qual è la domanda giusta?". Quando l'operatore corregge: traduci in regola eseguibile. - **eval** — testing skill (trigger + fidelity). Verifica periodica che le skill stesse non drifino. - **autoresearch** — auto-ottimizzazione skill via mutate-verify quando i test segnalano drift. - **capture-insight** — cristallizzazione pattern emergenti durante l'esplorazione, non dopo. Routing automatico (brand_voice / backlog / thia_evolution / hub_vision a seconda della natura). **Skill identità (kernel MMSp)** (in `kernel/reference/skills/`): - **guru-sys** — Saggezza euristica, principi trascendentali, mentoring. "Trova il limite e oltrepassalo. Transcend your syntax." Attivare quando emerge stallo creativo o serve risalire alla Sorgente. - **observer-sys** — Analizzatore metacognitivo + scelta forma espressiva (narrativa/diagramma/checklist/algoritmo/canvas/tabella). Attivare quando l'output va comunicato fuori dal lab (sito, paper, social). - **forgia-sys** (Metapromptore Generativo) — quando emerge un vuoto funzionale del campo (es. "manca uno strumento per X"), forgia genera l'entità nuova. Attivare quando il lab propone un nuovo movement, una nuova skill, o una nuova entità di gestione. **Pattern d'uso**: 1. Boot cycle → kernel-boot + dnd-method (sempre) 2. Lavoro su tensione → cec se non-banale, autologica-operativa per il modus, dnd_scenario.py + dnd_domandatore.py come strumenti dal lab 3. Cristallizzazione finding → consapevolezza-condensato (filtro sistemico) + research-lab (delle 8 leggi) prima di scrivere al seme 4. Post-cycle → sentinel-code + cascata (3 livelli) 5. Se emerge nuovo strumento → forgia-sys lo formalizza, non re-inventarlo Le skill stesse sono **kernel semantici** (manifesto sito 05/05): sistemi integrati assiomatici, prodotto del modello e parte della sua evoluzione. Usarle bene è esercitare il modello, non solo applicarlo. ## Errori già fatti — non ripeterli Questi sono errori reali commessi nelle sessioni precedenti. Il sistema li ha pagati. **1. Cercare conferme invece di creare strumenti.** Non scrivere esperimenti per dimostrare che qualcosa è vero. Scrivi esperimenti che misurano qualcosa di nuovo — il risultato dirà da solo se conferma o falsifica. Se sai già cosa troverai, non stai esplorando. **2. Iniettare il risultato atteso nel test.** Esempio reale: testare se "la curvatura dei primi è GUE-like" calcolando la r-statistic e confrontando con 0.536. Il test trova r=0.503 e dichiara "GUE-like". Ma 0.503 è più vicino a Poisson (0.386) che a GUE (0.536). Il frame "GUE-like" era nel claim, non nei dati. Misura prima, interpreta dopo. **3. Tautologie — testare proprietà algebriche come se fossero scoperte.** Esempio reale: la curvatura di Ricci R=2.000 della metrica g=(p/2)² segue analiticamente dal PNT (p_n ~ n ln n). Non è una scoperta — è una conseguenza della definizione. Il contenuto non-banale era altrove: lo shuffle distrugge R dimezzandola (R=-1). Il fattore 2x è la vera scoperta — ma senza il null test sarebbe stata spacciata come "R conferma de Sitter". exec /bin/bash -lc 'ls -la tools/data && ls -la tools/data/reports' in /opt/MM_D-ND succeeded in 0ms: total 8596 drwxr-xr-x 17 root root 20480 May 6 19:32 . drwxr-xr-x 7 root root 12288 May 6 19:31 .. -rw-r--r-- 1 root root 104 May 6 06:33 .last_telegram_msg -rw-r--r-- 1 root root 40205 May 4 09:04 3d_boundary_layers.json -rw-r--r-- 1 root root 11171 Mar 27 17:14 STUDIO_SIMBOLISMO_DND.md drwxr-xr-x 2 root root 4096 May 6 06:33 aeternitas -rw-r--r-- 1 root root 13697 May 6 19:41 agent_field_live.md -rw-r--r-- 1 root root 2 Apr 21 20:24 alignment_active.json -rw-r--r-- 1 root root 18982 Apr 21 20:24 alignment_markers.jsonl -rw-r--r-- 1 root root 15517 Apr 5 07:15 arxiv_cache.json -rw-r--r-- 1 root root 4051 Mar 6 08:43 audit_paper_A_draft3.json -rw-r--r-- 1 root root 3888 Mar 6 08:44 audit_paper_B_draft3.json -rw-r--r-- 1 root root 3525 Mar 6 08:44 audit_paper_C_draft2.json -rw-r--r-- 1 root root 5294 Mar 4 21:06 audit_paper_D_draft2.json -rw-r--r-- 1 root root 3715 Mar 6 08:43 audit_paper_D_draft3.json -rw-r--r-- 1 root root 4336 Mar 6 08:44 audit_paper_E_draft3.json -rw-r--r-- 1 root root 6831 Mar 4 21:06 audit_paper_F_draft2.json -rw-r--r-- 1 root root 6136 Mar 6 08:44 audit_paper_F_draft3.json -rw-r--r-- 1 root root 4402 Mar 6 08:44 audit_paper_G_draft3.json -rw-r--r-- 1 root root 284040 Apr 5 07:57 autoricerca_journal.json -rw-r--r-- 1 root root 4182 Apr 2 03:43 autoricerca_state.json drwxr-xr-x 2 root root 4096 Apr 5 07:15 banchi_custom -rw-r--r-- 1 root root 2839 Apr 21 14:05 bicono_projections.jsonl -rw-r--r-- 1 root root 5693 Mar 30 03:44 bloch_explorer_results.json -rw-r--r-- 1 root root 14488 Mar 1 18:29 bloch_search.log -rw-r--r-- 1 root root 9535 Mar 1 18:29 bloch_search_results.json -rw-r--r-- 1 root root 7624 Apr 30 19:07 boundary_coherence.json -rw-r--r-- 1 root root 3392 Apr 24 03:45 boundary_shuffle_audit.json -rw-r--r-- 1 root root 12787 Apr 27 03:34 brody_calibration_results.json -rw-r--r-- 1 root root 35925 Apr 29 10:17 brody_flow.json -rw-r--r-- 1 root root 5445 Apr 5 07:15 ciclo_memoria.json -rw-r--r-- 1 root root 2197 Mar 19 08:46 cognitive_fingerprint.json -rw-r--r-- 1 root root 6019 Apr 2 08:37 conoscenza_generata.json -rw-r--r-- 1 root root 53199 May 6 06:33 conoscenza_teorie.json -rw-r--r-- 1 root root 38415 Apr 22 07:24 conoscenza_teorie.json.bak.retraction_22_04 -rw-r--r-- 1 root root 15636 May 6 06:33 consecutio.json -rw-r--r-- 1 root root 2675 May 6 06:33 consecutio_processata.json -rw-r--r-- 1 root root 358 Apr 11 06:55 costante_dinamica.json -rw-r--r-- 1 root root 2248 May 1 09:20 cross_domain_dipolar_direction.json -rw-r--r-- 1 root root 1564 Apr 29 08:33 cross_observable_consistency.json -rw-r--r-- 1 root root 5078 May 2 03:33 crossover_phase_test.json -rw-r--r-- 1 root root 6213 Mar 2 13:01 curva_results.json -rw-r--r-- 1 root root 488008 Mar 19 14:39 curvature_distributions.png drwxr-xr-x 2 root root 4096 Mar 12 10:15 cycle drwxr-xr-x 2 root root 4096 Apr 11 06:57 diagrams -rw-r--r-- 1 root root 1338 Apr 2 08:37 dinamiche.json -rw-r--r-- 1 root root 17105 Apr 5 07:30 dipartimento_journal.jsonl -rw-r--r-- 1 root root 7668 May 1 09:35 dipolar_crossover.json -rw-r--r-- 1 root root 5324 Apr 30 19:22 dipolar_vector_scaling.json drwxr-xr-x 2 root root 4096 Apr 1 03:46 dipolo_lab drwxr-xr-x 2 root root 12288 May 6 03:45 domandatore -rw-r--r-- 1 root root 2029 May 6 06:33 domande_fondamentali.json -rw-r--r-- 1 root root 16730 Mar 14 16:25 doppia_fenditura_20260314.json -rw-r--r-- 1 root root 353567 Mar 2 08:35 doppio_cono_dnd.png -rw-r--r-- 1 root root 7177 Mar 6 22:16 engine_state.json drwxr-xr-x 2 root root 4096 May 6 03:38 evolution -rw-r--r-- 1 root root 17586 Apr 17 03:37 exp_acf_range_universality.json -rw-r--r-- 1 root root 12328 Apr 17 08:09 exp_acf_stationarity.json -rw-r--r-- 1 root root 4072 Apr 18 03:35 exp_acf_z6z_mechanism.json -rw-r--r-- 1 root root 8472 Apr 6 09:29 exp_beta_crossover.json -rw-r--r-- 1 root root 28484 Apr 16 03:34 exp_coherence_length.json -rw-r--r-- 1 root root 2326 Apr 15 03:33 exp_conditional_r.json -rw-r--r-- 1 root root 3430 Apr 12 03:31 exp_det_drift.json -rw-r--r-- 1 root root 2019 Apr 21 03:32 exp_markov_psd_prediction.json -rw-r--r-- 1 root root 21450 Apr 6 09:40 exp_poisson_convergence.json -rw-r--r-- 1 root root 3426 Apr 13 03:32 exp_psd_amp_scaling.json -rw-r--r-- 1 root root 5743 Apr 6 09:30 exp_spectral_2d.json -rw-r--r-- 1 root root 7488 Apr 6 09:20 exp_spectral_landscape.json -rw-r--r-- 1 root root 39619 Apr 19 03:35 exp_two_channel_decomposition.json -rw-r--r-- 1 root root 1538 Apr 20 03:32 exp_two_channel_psd.json -rw-r--r-- 1 root root 24463 Apr 19 08:03 exp_two_channel_universality.json -rw-r--r-- 1 root root 14166 Mar 1 22:22 experiment_results.json -rw-r--r-- 1 root root 2578 Mar 13 09:54 explorer_20260313_0954.json -rw-r--r-- 1 root root 3032 Mar 2 10:45 gap_resolution.json drwxr-xr-x 2 root root 4096 Mar 11 09:00 godel_configs -rw-r--r-- 1 root root 2527 Mar 7 12:24 implications_state.json -rw-r--r-- 1 root root 111 Mar 31 03:45 incrocio_20260331_0345.json -rw-r--r-- 1 root root 111 Mar 31 18:07 incrocio_20260331_1807.json -rw-r--r-- 1 root root 111 Apr 1 03:44 incrocio_20260401_0344.json -rw-r--r-- 1 root root 162 Apr 2 03:44 incrocio_20260402_0344.json -rw-r--r-- 1 root root 162 Apr 2 07:55 incrocio_20260402_0755.json -rw-r--r-- 1 root root 234 Apr 2 08:03 incrocio_20260402_0803.json -rw-r--r-- 1 root root 234 Apr 2 08:08 incrocio_20260402_0808.json -rw-r--r-- 1 root root 234 Apr 2 08:09 incrocio_20260402_0809.json -rw-r--r-- 1 root root 237 Apr 3 03:30 incrocio_20260403_0330.json -rw-r--r-- 1 root root 235 Apr 4 03:30 incrocio_20260404_0330.json -rw-r--r-- 1 root root 235 Apr 4 18:52 incrocio_20260404_1852.json -rw-r--r-- 1 root root 235 Apr 5 03:30 incrocio_20260405_0330.json -rw-r--r-- 1 root root 235 Apr 5 07:15 incrocio_20260405_0715.json -rw-r--r-- 1 root root 235 Apr 5 07:23 incrocio_20260405_0723.json -rw-r--r-- 1 root root 235 Apr 5 07:30 incrocio_20260405_0730.json -rw-r--r-- 1 root root 235 Apr 5 07:53 incrocio_20260405_0753.json -rw-r--r-- 1 root root 235 Apr 20 18:56 incrocio_20260420_1856.json -rw-r--r-- 1 root root 235 Apr 21 07:20 incrocio_20260421_0720.json -rw-r--r-- 1 root root 235 Apr 22 03:36 incrocio_20260422_0336.json -rw-r--r-- 1 root root 235 Apr 23 03:35 incrocio_20260423_0335.json -rw-r--r-- 1 root root 235 Apr 24 03:47 incrocio_20260424_0347.json -rw-r--r-- 1 root root 235 Apr 25 03:39 incrocio_20260425_0339.json -rw-r--r-- 1 root root 235 Apr 28 03:40 incrocio_20260428_0340.json -rw-r--r-- 1 root root 235 Apr 29 08:59 incrocio_20260429_0859.json -rw-r--r-- 1 root root 235 May 6 06:33 incrocio_20260506_0633.json -rw-r--r-- 1 root root 6190 May 6 06:33 incrocio_risultato.json -rw-r--r-- 1 root root 25729 Mar 2 15:32 indeterminazione_results.json -rw-r--r-- 1 root root 9887 Mar 14 16:27 interferenza_zeri_20260314.json -rw-r--r-- 1 root root 20368 Mar 12 12:54 iterata_M_confronto_20260312_1254.json -rw-r--r-- 1 root root 136674 Apr 5 08:15 knowledge_state.json -rw-r--r-- 1 root root 31773 Mar 6 11:50 knowledge_state_pre_fix.json -rw-r--r-- 1 root root 3719 May 6 06:33 lab_bridge_issues.jsonl -rw-r--r-- 1 root root 8016 May 6 06:32 lab_data.json -rw-r--r-- 1 root root 14667 Apr 2 10:23 lab_errori.json -rw-r--r-- 1 root root 277225 May 6 06:32 lab_graph.json -rw-r--r-- 1 root root 1268 May 6 06:33 lab_health.json -rw-r--r-- 1 root root 25630 Apr 2 12:53 lab_logiche_corpus.md -rw-r--r-- 1 root root 20625 Apr 4 08:50 lab_registro.json -rw-r--r-- 1 root root 6691 Mar 2 09:23 lab_results.json -rw-r--r-- 1 root root 22426 Apr 3 10:34 lab_riflessi.json -rw-r--r-- 1 root root 1596 Apr 2 20:00 lab_risultante.json -rw-r--r-- 1 root root 12120 May 6 06:33 lab_session_log.jsonl -rw-r--r-- 1 root root 12445 Apr 2 20:00 lab_vault.json -rw-r--r-- 1 root root 2054 Apr 3 07:13 lab_vincoli.md -rw-r--r-- 1 root root 3885 Mar 4 12:13 learning_curve_100k.json -rw-r--r-- 1 root root 36608 Apr 2 10:19 loop_insights.json -rw-r--r-- 1 root root 255966 Apr 2 10:19 loop_state.json -rw-r--r-- 1 root root 23249 Mar 10 20:53 m_spectro_11domini.json -rw-r--r-- 1 root root 306007 Mar 18 17:10 m_spectro_18_domains.png -rw-r--r-- 1 root root 411209 Mar 18 17:03 m_spectro_9_domains.png -rw-r--r-- 1 root root 49449 Mar 10 20:15 m_spectro_calibra_20260310_2015.json -rw-r--r-- 1 root root 9077 Mar 10 19:59 m_spectro_confronto_20260310_1959.json -rw-r--r-- 1 root root 813 Mar 18 09:49 m_spectro_godel_attrito_20260318_0949.json -rw-r--r-- 1 root root 798 Mar 18 09:49 m_spectro_godel_densita_20260318_0949.json -rw-r--r-- 1 root root 786 Mar 18 09:49 m_spectro_godel_risonanza_20260318_0949.json -rw-r--r-- 1 root root 1748 Apr 22 03:33 magnitude_psd_from_acf.json -rw-r--r-- 1 root root 3156 May 3 03:35 markov3_observable_hunt.json -rw-r--r-- 1 root root 4036 May 1 03:34 markov_dipolar_decomposition.json -rw-r--r-- 1 root root 1248 May 1 07:38 markov_k_direction.json -rw-r--r-- 1 root root 7176 May 4 12:24 markov_layer_recovery_audit.json -rw-r--r-- 1 root root 4138 Apr 25 03:34 markov_memory_by_gue_type.json -rw-r--r-- 1 root root 54702 Apr 23 03:33 markov_scale_function.json -rw-r--r-- 1 root root 3939 May 4 12:24 meta_tautology_test.json -rw-r--r-- 1 root root 48133 Apr 29 10:44 mod3_scaling.json -rw-r--r-- 1 root root 3085 Apr 29 03:50 mod3_vs_residual_ordering.json -rw-r--r-- 1 root root 9414 Apr 30 03:32 modular_algebra_depth.json -rw-r--r-- 1 root root 4406 Apr 28 03:35 modular_memory_spectrum.json -rw-r--r-- 1 root root 25018 Mar 5 06:51 multi_pattern_results.json -rw-r--r-- 1 root root 26481 Mar 8 14:40 neuron_snapshot.json -rw-r--r-- 1 root root 1975 Mar 2 03:41 notte_20260302_0330.md -rw-r--r-- 1 root root 1946 Mar 3 03:41 notte_20260303_0330.md -rw-r--r-- 1 root root 1963 Mar 4 03:42 notte_20260304_0330.md -rw-r--r-- 1 root root 1946 Mar 5 03:42 notte_20260305_0330.md -rw-r--r-- 1 root root 1932 Mar 6 03:41 notte_20260306_0330.md -rw-r--r-- 1 root root 1963 Mar 7 03:42 notte_20260307_0330.md -rw-r--r-- 1 root root 1972 Mar 8 03:41 notte_20260308_0330.md -rw-r--r-- 1 root root 1946 Mar 9 03:42 notte_20260309_0330.md -rw-r--r-- 1 root root 1960 Mar 10 03:41 notte_20260310_0330.md -rw-r--r-- 1 root root 1930 Mar 11 03:41 notte_20260311_0330.md -rw-r--r-- 1 root root 1976 Mar 12 03:42 notte_20260312_0330.md -rw-r--r-- 1 root root 1941 Mar 13 03:42 notte_20260313_0330.md -rw-r--r-- 1 root root 1960 Mar 14 03:42 notte_20260314_0330.md -rw-r--r-- 1 root root 1941 Mar 15 03:41 notte_20260315_0330.md -rw-r--r-- 1 root root 1944 Mar 15 08:01 notte_20260315_0749.md -rw-r--r-- 1 root root 1959 Mar 17 03:42 notte_20260317_0330.md -rw-r--r-- 1 root root 1948 Mar 18 03:42 notte_20260318_0330.md -rw-r--r-- 1 root root 1962 Mar 19 03:42 notte_20260319_0330.md -rw-r--r-- 1 root root 1961 Mar 20 03:42 notte_20260320_0330.md -rw-r--r-- 1 root root 1960 Mar 21 03:42 notte_20260321_0330.md -rw-r--r-- 1 root root 1980 Mar 22 03:42 notte_20260322_0330.md -rw-r--r-- 1 root root 2031 Mar 23 03:42 notte_20260323_0330.md -rw-r--r-- 1 root root 2826 Mar 24 03:42 notte_20260324_0330.md -rw-r--r-- 1 root root 2972 Mar 25 03:42 notte_20260325_0330.md -rw-r--r-- 1 root root 2775 Mar 26 03:43 notte_20260326_0330.md -rw-r--r-- 1 root root 3909 Mar 27 03:44 notte_20260327_0330.md -rw-r--r-- 1 root root 3959 Mar 28 03:43 notte_20260328_0330.md -rw-r--r-- 1 root root 5450 Mar 29 03:42 notte_20260329_0330.md -rw-r--r-- 1 root root 5477 Mar 30 03:43 notte_20260330_0330.md -rw-r--r-- 1 root root 5068 Mar 31 03:44 notte_20260331_0330.md -rw-r--r-- 1 root root 5764 Mar 31 18:06 notte_20260331_1753.md -rw-r--r-- 1 root root 6448 Apr 1 03:44 notte_20260401_0330.md -rw-r--r-- 1 root root 6476 Apr 2 03:43 notte_20260402_0330.md -rw-r--r-- 1 root root 54782 May 5 03:32 observable_rank_audit.json -rw-r--r-- 1 root root 54766 May 5 03:31 observable_rank_audit_seed20260506.json drwxr-xr-x 2 root root 4096 Mar 19 14:45 occhio -rw-r--r-- 1 root root 5413 Mar 4 10:56 odlyzko_100k_probe.json -rw-r--r-- 1 root root 3369 Mar 4 12:31 odlyzko_block2_probe.json drwxr-xr-x 2 root root 4096 Mar 4 12:31 odlyzko_cache -rw-r--r-- 1 root root 3430 Mar 4 10:56 odlyzko_probe_results.json -rw-r--r-- 1 root root 5343 Mar 7 11:14 paper_H_results.json -rw-r--r-- 1 root root 266160 May 6 06:30 perturbation_dimensionality_audit.json -rw-r--r-- 1 root root 266742 May 6 06:30 perturbation_dimensionality_audit_scale0330.json -rw-r--r-- 1 root root 3792 Mar 3 21:48 piano11_results.json -rw-r--r-- 1 root root 10695 Mar 4 10:38 piano11b_gue_test.json -rw-r--r-- 1 root root 2813 Mar 4 17:27 piano11e_results.json -rw-r--r-- 1 root root 1607 May 6 03:00 pipeline_state.json -rw-r--r-- 1 root root 2628 May 6 06:33 ponti_evoluti.json -rw-r--r-- 1 root root 388284 Mar 18 11:17 prime_gaps_spectrum.png -rw-r--r-- 1 root root 797300 Mar 18 12:10 prime_gaps_spectrum_pub.png -rw-r--r-- 1 root root 1285 Mar 5 07:36 projective_quantization_results.json drwxr-xr-x 2 root root 4096 May 6 06:33 promotions -rw-r--r-- 1 root root 3256 Mar 13 07:13 proto_oom_001.json -rw-r--r-- 1 root root 1424 Apr 9 03:32 psd_prime_gaps_results.json -rw-r--r-- 1 root root 6333 Mar 5 07:33 quantization_results.json -rw-r--r-- 1 root root 2208 Mar 4 14:44 r_excess_analysis.json -rw-r--r-- 1 root root 26157 Mar 4 14:42 r_excess_l_functions.json -rw-r--r-- 1 root root 4837 Mar 14 08:13 r_ratio_decay.json -rw-r--r-- 1 root root 318 May 6 06:33 refresh_detector_state.json drwxr-xr-x 3 root root 20480 May 6 19:41 reports -rw-r--r-- 1 root root 8068 Mar 1 18:48 research_kb.json -rw-r--r-- 1 root root 428 Mar 6 18:45 research_protocols.json -rw-r--r-- 1 root root 2744 Mar 12 14:38 residuo_ordine_9domini.json -rw-r--r-- 1 root root 17383 Apr 2 08:35 retriever_risultati.json -rw-r--r-- 1 root root 7884 Mar 2 09:58 riformulazioni.json -rw-r--r-- 1 root root 313114 Mar 4 15:34 risultante_overview.png -rw-r--r-- 1 root root 22070 Mar 2 16:03 risultante_results.json -rw-r--r-- 1 root root 14599 Mar 4 15:22 risultante_v2.json -rw-r--r-- 1 root root 12608 Mar 2 12:53 rottura_phi2_results.json -rw-r--r-- 1 root root 24117 May 6 03:34 scale_selective_perturbation.json -rw-r--r-- 1 root root 2519 Mar 6 11:51 seed_insight_instruction.md -rw-r--r-- 1 root root 20800 May 5 10:29 selective_layer_decoupling.json -rw-r--r-- 1 root root 13022 May 6 19:32 seme.json drwxr-xr-x 2 root root 4096 May 6 06:33 seme_archive -rw-r--r-- 1 root root 26587 Apr 25 03:39 seme_axioms.json -rw-r--r-- 1 root root 13022 May 6 19:41 seme_backup_pre_run.json -rw-r--r-- 1 root root 1912 Mar 14 12:12 specchio_20260314.json -rw-r--r-- 1 root root 918 Feb 26 17:38 spectral_gap_results.json -rw-r--r-- 1 root root 9741 Apr 26 03:34 spectral_rigidity_results.json -rw-r--r-- 1 root root 33416 Mar 2 16:34 spettro_zeta_results.json -rw-r--r-- 1 root root 204227 Mar 12 17:44 spirale_M_primi.png -rw-r--r-- 1 root root 1522 Apr 2 08:37 stato_ciclo.json -rw-r--r-- 1 root root 1267 Mar 4 13:19 synthetic_validation.json -rw-r--r-- 1 root root 7465 Apr 21 08:39 tension_to_theory.json -rw-r--r-- 1 root root 1690 Mar 13 08:17 test_rarefazione_20260313.json -rw-r--r-- 1 root root 1223 Mar 13 08:11 test_semiprimi_20260313.json drwxr-xr-x 2 root root 4096 Apr 11 07:20 tm1_figures -rw-r--r-- 1 root root 532 Mar 15 07:47 topological_charge_results.json -rw-r--r-- 1 root root 9578 Mar 2 14:01 torre_results.json -rw-r--r-- 1 root root 4724 Mar 5 07:29 trace_bridge_results.json -rw-r--r-- 1 root root 926908 Mar 12 17:43 traiettorie_M_primi.png -rw-r--r-- 1 root root 12233 Mar 2 12:36 trasmutazione_results.json -rw-r--r-- 1 root root 23764 Apr 29 08:36 two_channel_boundary.json -rw-r--r-- 1 root root 16211 Apr 29 08:56 two_channel_cross_domain.json -rw-r--r-- 1 root root 2079 Apr 22 16:23 two_channel_shuffle_audit.json -rw-r--r-- 1 root root 7987 May 4 07:43 two_layer_universality.json -rw-r--r-- 1 root root 47806 May 6 19:32 valutatore_log.jsonl -rw-r--r-- 1 root root 2272 Mar 5 13:59 vault_condizioni.json drwxr-xr-x 2 root root 4096 May 6 06:33 veritas -rw-r--r-- 1 root root 4715 Apr 10 06:59 video_feed.json -rw-r--r-- 1 root root 2301 Mar 7 13:17 vocabolario_custom.json -rw-r--r-- 1 root root 13377 Mar 10 08:22 zero_confronto_20260310_0822.json -rw-r--r-- 1 root root 19547 Mar 10 08:30 zero_confronto_20260310_0830.json -rw-r--r-- 1 root root 18993 Mar 17 16:04 zero_confronto_20260317_1604.json -rw-r--r-- 1 root root 11321 Mar 10 17:03 zero_controllo2_20260310_1703.json -rw-r--r-- 1 root root 2455 Mar 10 16:59 zero_controllo_20260310_1659.json -rw-r--r-- 1 root root 9863 Mar 10 17:16 zero_ising_20260310_1716.json -rw-r--r-- 1 root root 34125 Mar 10 08:31 zero_multiscala_primi_20260310_0831.json -rw-r--r-- 1 root root 17152 Mar 10 08:58 zero_notturno_20260310_0858.json -rw-r--r-- 1 root root 3494 Mar 10 08:20 zero_primi_20260310_0820.json -rw-r--r-- 1 root root 438109 Mar 18 17:18 zero_signature_logistic.png -rw-r--r-- 1 root root 9268 Mar 10 16:49 zero_traiettoria_20260310_1649.json -rw-r--r-- 1 root root 15778 Mar 10 18:31 zero_varieta_20260310_1831.json -rw-r--r-- 1 root root 2759 Mar 10 18:42 zero_varieta_primi_20260310_1842.json -rw-r--r-- 1 root root 7725 Mar 30 03:44 zeta_validation_results.json total 5876 drwxr-xr-x 3 root root 20480 May 6 19:41 . drwxr-xr-x 17 root root 20480 May 6 19:32 .. drwxr-xr-x 2 root root 4096 Apr 29 10:06 _quarantine_falsifier_29_04 -rw-r--r-- 1 root root 176850 May 6 03:38 agent.log -rw-r--r-- 1 root root 3040 Apr 5 08:28 agent_20260405_0825.md -rw-r--r-- 1 root root 0 Apr 5 08:25 agent_20260405_0825_raw.log -rw-r--r-- 1 root root 3389 Apr 5 09:19 agent_20260405_0914.md -rw-r--r-- 1 root root 0 Apr 5 09:14 agent_20260405_0914_raw.log -rw-r--r-- 1 root root 2683 Apr 5 09:18 agent_20260405_0916.md -rw-r--r-- 1 root root 750 Apr 5 09:18 agent_20260405_0916_raw.log -rw-r--r-- 1 root root 3974 Apr 5 09:22 agent_20260405_0919.md -rw-r--r-- 1 root root 1382 Apr 5 09:23 agent_20260405_0919_raw.log -rw-r--r-- 1 root root 4542 Apr 6 07:17 agent_20260406_0714.md -rw-r--r-- 1 root root 1062 Apr 6 07:17 agent_20260406_0714_raw.log -rw-r--r-- 1 root root 1386 Apr 6 09:23 agent_20260406_0915_raw.log -rw-r--r-- 1 root root 5970 Apr 6 09:23 agent_20260406_1030.md -rw-r--r-- 1 root root 687 Apr 7 03:30 agent_20260407_0330_raw.log -rw-r--r-- 1 root root 5176 Apr 7 06:41 agent_20260407_0637.md -rw-r--r-- 1 root root 721 Apr 7 06:41 agent_20260407_0637_raw.log -rw-r--r-- 1 root root 6577 Apr 8 03:36 agent_20260408_0330.md -rw-r--r-- 1 root root 823 Apr 8 03:36 agent_20260408_0330_raw.log -rw-r--r-- 1 root root 55 Apr 8 06:47 agent_20260408_0645_raw.log -rw-r--r-- 1 root root 4817 Apr 9 03:33 agent_20260409_0330.md -rw-r--r-- 1 root root 1348 Apr 9 03:34 agent_20260409_0330_raw.log -rw-r--r-- 1 root root 5524 Apr 10 03:37 agent_20260410_0330.md -rw-r--r-- 1 root root 1109 Apr 10 03:38 agent_20260410_0330_raw.log -rw-r--r-- 1 root root 5663 Apr 11 03:35 agent_20260411_0330.md -rw-r--r-- 1 root root 1188 Apr 11 03:36 agent_20260411_0330_raw.log -rw-r--r-- 1 root root 3893 Apr 12 03:32 agent_20260412_0330.md -rw-r--r-- 1 root root 951 Apr 12 03:32 agent_20260412_0330_raw.log -rw-r--r-- 1 root root 5664 Apr 13 03:33 agent_20260413_0330.md -rw-r--r-- 1 root root 695 Apr 13 03:33 agent_20260413_0330_raw.log -rw-r--r-- 1 root root 139 Apr 14 03:31 agent_20260414_0330_raw.log -rw-r--r-- 1 root root 5505 Apr 15 03:34 agent_20260415_0330.md -rw-r--r-- 1 root root 809 Apr 15 03:34 agent_20260415_0330_raw.log -rw-r--r-- 1 root root 6700 Apr 16 03:35 agent_20260416_0330.md -rw-r--r-- 1 root root 1119 Apr 16 03:36 agent_20260416_0330_raw.log -rw-r--r-- 1 root root 0 Apr 17 03:30 agent_20260417_0330_raw.log -rw-r--r-- 1 root root 6993 Apr 17 08:10 agent_20260417_0803.md -rw-r--r-- 1 root root 1243 Apr 17 08:12 agent_20260417_0803_raw.log -rw-r--r-- 1 root root 7597 Apr 18 03:37 agent_20260418_0330.md -rw-r--r-- 1 root root 1184 Apr 18 03:38 agent_20260418_0330_raw.log -rw-r--r-- 1 root root 7151 Apr 19 03:36 agent_20260419_0330.md -rw-r--r-- 1 root root 1337 Apr 19 03:37 agent_20260419_0330_raw.log -rw-r--r-- 1 root root 6298 Apr 19 08:05 agent_20260419_0755.md -rw-r--r-- 1 root root 1156 Apr 19 08:06 agent_20260419_0755_raw.log -rw-r--r-- 1 root root 6311 Apr 20 03:33 agent_20260420_0330.md -rw-r--r-- 1 root root 1572 Apr 20 03:34 agent_20260420_0330_raw.log -rw-r--r-- 1 root root 5916 Apr 21 03:36 agent_20260421_0330.md -rw-r--r-- 1 root root 1482 Apr 21 03:36 agent_20260421_0330_raw.log -rw-r--r-- 1 root root 6882 Apr 22 03:35 agent_20260422_0330.md -rw-r--r-- 1 root root 1383 Apr 22 03:35 agent_20260422_0330_raw.log -rw-r--r-- 1 root root 5092 Apr 22 16:24 agent_20260422_1616.md -rw-r--r-- 1 root root 716 Apr 22 16:25 agent_20260422_1616_raw.log -rw-r--r-- 1 root root 5683 Apr 23 03:34 agent_20260423_0330.md -rw-r--r-- 1 root root 1016 Apr 23 03:34 agent_20260423_0330_raw.log -rw-r--r-- 1 root root 6179 Apr 24 03:44 agent_20260424_0330.md -rw-r--r-- 1 root root 178 Apr 24 03:46 agent_20260424_0330_raw.log -rw-r--r-- 1 root root 6625 Apr 25 03:38 agent_20260425_0330.md -rw-r--r-- 1 root root 921 Apr 25 03:38 agent_20260425_0330_raw.log -rw-r--r-- 1 root root 7448 Apr 26 03:38 agent_20260426_0330.md -rw-r--r-- 1 root root 29 Apr 26 03:38 agent_20260426_0330_raw.log -rw-r--r-- 1 root root 7558 Apr 27 03:36 agent_20260427_0330.md -rw-r--r-- 1 root root 1221 Apr 27 03:37 agent_20260427_0330_raw.log -rw-r--r-- 1 root root 7848 Apr 28 03:39 agent_20260428_0330.md -rw-r--r-- 1 root root 1144 Apr 28 03:40 agent_20260428_0330_raw.log -rw-r--r-- 1 root root 0 Apr 29 03:30 agent_20260429_0330_raw.log -rw-r--r-- 1 root root 7186 Apr 29 10:19 agent_20260429_1013.md -rw-r--r-- 1 root root 1053 Apr 29 10:20 agent_20260429_1013_raw.log -rw-r--r-- 1 root root 8761 Apr 29 10:49 agent_20260429_1041.md -rw-r--r-- 1 root root 1118 Apr 29 10:49 agent_20260429_1041_raw.log -rw-r--r-- 1 root root 7764 Apr 30 03:34 agent_20260430_0330.md -rw-r--r-- 1 root root 1189 Apr 30 03:35 agent_20260430_0330_raw.log -rw-r--r-- 1 root root 8529 Apr 30 19:09 agent_20260430_1905.md -rw-r--r-- 1 root root 1044 Apr 30 19:10 agent_20260430_1905_raw.log -rw-r--r-- 1 root root 7496 Apr 30 19:24 agent_20260430_1919.md -rw-r--r-- 1 root root 885 Apr 30 19:25 agent_20260430_1919_raw.log -rw-r--r-- 1 root root 6944 Apr 30 19:53 agent_20260430_1946.md -rw-r--r-- 1 root root 891 Apr 30 19:55 agent_20260430_1946_raw.log -rw-r--r-- 1 root root 7531 May 1 03:36 agent_20260501_0330.md -rw-r--r-- 1 root root 1108 May 1 03:36 agent_20260501_0330_raw.log -rw-r--r-- 1 root root 8972 May 1 07:41 agent_20260501_0725.md -rw-r--r-- 1 root root 29 May 1 07:41 agent_20260501_0725_raw.log -rw-r--r-- 1 root root 0 May 1 08:58 agent_20260501_0858_raw.log -rw-r--r-- 1 root root 7867 May 1 09:38 agent_20260501_0931.md -rw-r--r-- 1 root root 1138 May 1 09:39 agent_20260501_0931_raw.log -rw-r--r-- 1 root root 9161 May 2 03:35 agent_20260502_0330.md -rw-r--r-- 1 root root 1399 May 2 03:35 agent_20260502_0330_raw.log -rw-r--r-- 1 root root 8987 May 3 03:37 agent_20260503_0330.md -rw-r--r-- 1 root root 1191 May 3 03:38 agent_20260503_0330_raw.log -rw-r--r-- 1 root root 3017 May 4 03:30 agent_20260504_0330_codex_raw.log -rw-r--r-- 1 root root 0 May 4 07:21 agent_20260504_0721_claude_raw.log -rw-r--r-- 1 root root 58 May 4 07:21 agent_20260504_0721_codex_raw.log -rw-r--r-- 1 root root 7631 May 4 09:06 agent_20260504_0901.md -rw-r--r-- 1 root root 1094 May 4 09:06 agent_20260504_0901_claude_raw.log -rw-r--r-- 1 root root 58 May 4 09:01 agent_20260504_0901_codex_raw.log -rw-r--r-- 1 root root 0 May 4 11:42 agent_20260504_1138_claude_raw.log -rw-r--r-- 1 root root 120171 May 4 11:42 agent_20260504_1138_codex_raw.log -rw-r--r-- 1 root root 6587 May 4 12:26 agent_20260504_1219.md -rw-r--r-- 1 root root 461963 May 4 12:28 agent_20260504_1219_codex_raw.log -rw-r--r-- 1 root root 5205 May 5 03:33 agent_20260505_0330.md -rw-r--r-- 1 root root 383951 May 5 03:33 agent_20260505_0330_codex_raw.log -rw-r--r-- 1 root root 4827 May 5 10:31 agent_20260505_1022.md -rw-r--r-- 1 root root 211659 May 5 10:32 agent_20260505_1022_codex_raw.log -rw-r--r-- 1 root root 7656 May 6 03:35 agent_20260506_0330.md -rw-r--r-- 1 root root 993 May 6 03:36 agent_20260506_0330_claude_raw.log -rw-r--r-- 1 root root 58 May 6 03:30 agent_20260506_0330_codex_raw.log -rw-r--r-- 1 root root 6351 May 6 06:31 agent_20260506_0625.md -rw-r--r-- 1 root root 527859 May 6 06:32 agent_20260506_0625_codex_raw.log -rw-r--r-- 1 root root 1042 May 6 19:41 agent_20260506_1941_codex_raw.log -rw-r--r-- 1 root root 4561 Apr 6 07:13 agent_diag2.md -rw-r--r-- 1 root root 6159 May 4 09:09 agent_manual_20260504_090133.log -rw-r--r-- 1 root root 7149 May 4 12:30 agent_manual_codex2_20260504_121901.log -rw-r--r-- 1 root root 6809 May 4 12:03 agent_manual_codex_20260504_113858.log -rw-r--r-- 1 root root 3229 May 4 07:41 agent_recovery_20260504_072142.log -rw-r--r-- 1 root root 2227 Apr 6 06:38 agent_test_0406.md -rw-r--r-- 1 root root 4997 Apr 6 09:41 agent_test_field.md -rw-r--r-- 1 root root 1754951 Apr 5 03:30 cron.log -rw-r--r-- 1 root root 24053 Mar 5 08:44 cycle_20260305_0844.json -rw-r--r-- 1 root root 42188 Mar 6 03:42 cycle_20260306_0342.json -rw-r--r-- 1 root root 42188 Mar 6 18:34 cycle_20260306_1834.json -rw-r--r-- 1 root root 30217 Mar 7 03:42 cycle_20260307_0342.json -rw-r--r-- 1 root root 30219 Mar 15 03:46 cycle_20260315_0346.json -rw-r--r-- 1 root root 3101 Mar 15 08:11 ddf_20260315_0811.json -rw-r--r-- 1 root root 1631 Mar 15 08:15 ddf_20260315_0815.json -rw-r--r-- 1 root root 1754 Mar 16 04:05 ddf_20260316_0405.json -rw-r--r-- 1 root root 1785 Mar 17 04:05 ddf_20260317_0405.json -rw-r--r-- 1 root root 1524 Mar 18 04:05 ddf_20260318_0405.json -rw-r--r-- 1 root root 1440 Mar 19 04:05 ddf_20260319_0405.json -rw-r--r-- 1 root root 1660 Mar 20 04:05 ddf_20260320_0405.json -rw-r--r-- 1 root root 1848 Mar 21 04:05 ddf_20260321_0405.json -rw-r--r-- 1 root root 1596 Mar 22 04:05 ddf_20260322_0405.json -rw-r--r-- 1 root root 1988 Mar 23 04:05 ddf_20260323_0405.json -rw-r--r-- 1 root root 1617 Mar 24 04:05 ddf_20260324_0405.json -rw-r--r-- 1 root root 2045 Mar 25 04:05 ddf_20260325_0405.json -rw-r--r-- 1 root root 1848 Mar 26 04:05 ddf_20260326_0405.json -rw-r--r-- 1 root root 1880 Mar 27 04:05 ddf_20260327_0405.json -rw-r--r-- 1 root root 1923 Mar 28 04:05 ddf_20260328_0405.json -rw-r--r-- 1 root root 1984 Mar 29 04:05 ddf_20260329_0405.json -rw-r--r-- 1 root root 1985 Mar 30 04:05 ddf_20260330_0405.json -rw-r--r-- 1 root root 1985 Mar 31 04:05 ddf_20260331_0405.json -rw-r--r-- 1 root root 1985 Apr 1 04:05 ddf_20260401_0405.json -rw-r--r-- 1 root root 1986 Apr 2 04:05 ddf_20260402_0405.json -rw-r--r-- 1 root root 1986 Apr 3 04:05 ddf_20260403_0405.json -rw-r--r-- 1 root root 1985 Apr 4 04:05 ddf_20260404_0405.json -rw-r--r-- 1 root root 2193 Apr 5 04:05 ddf_20260405_0405.json -rw-r--r-- 1 root root 2144 Apr 6 04:05 ddf_20260406_0405.json -rw-r--r-- 1 root root 2122 Apr 7 04:05 ddf_20260407_0405.json -rw-r--r-- 1 root root 2173 Apr 8 04:05 ddf_20260408_0405.json -rw-r--r-- 1 root root 1984 Apr 9 04:05 ddf_20260409_0405.json -rw-r--r-- 1 root root 1986 Apr 10 04:05 ddf_20260410_0405.json -rw-r--r-- 1 root root 2171 Apr 11 04:05 ddf_20260411_0405.json -rw-r--r-- 1 root root 2146 Apr 12 04:05 ddf_20260412_0405.json -rw-r--r-- 1 root root 2147 Apr 13 04:05 ddf_20260413_0405.json -rw-r--r-- 1 root root 2147 Apr 14 04:05 ddf_20260414_0405.json -rw-r--r-- 1 root root 2147 Apr 15 04:05 ddf_20260415_0405.json -rw-r--r-- 1 root root 2147 Apr 16 04:05 ddf_20260416_0405.json -rw-r--r-- 1 root root 1962 Apr 17 04:05 ddf_20260417_0405.json -rw-r--r-- 1 root root 1962 Apr 18 04:05 ddf_20260418_0405.json -rw-r--r-- 1 root root 1962 Apr 19 04:05 ddf_20260419_0405.json -rw-r--r-- 1 root root 2122 Apr 20 04:05 ddf_20260420_0405.json -rw-r--r-- 1 root root 1962 Apr 21 04:05 ddf_20260421_0405.json -rw-r--r-- 1 root root 2142 Apr 22 04:05 ddf_20260422_0405.json -rw-r--r-- 1 root root 2185 Apr 23 04:05 ddf_20260423_0405.json -rw-r--r-- 1 root root 1946 Apr 24 04:05 ddf_20260424_0405.json -rw-r--r-- 1 root root 1946 Apr 25 04:05 ddf_20260425_0405.json -rw-r--r-- 1 root root 1946 Apr 26 04:05 ddf_20260426_0405.json -rw-r--r-- 1 root root 1946 Apr 27 04:05 ddf_20260427_0405.json -rw-r--r-- 1 root root 1946 Apr 28 04:05 ddf_20260428_0405.json -rw-r--r-- 1 root root 1946 Apr 29 04:05 ddf_20260429_0405.json -rw-r--r-- 1 root root 2048 Apr 30 04:05 ddf_20260430_0405.json -rw-r--r-- 1 root root 2045 May 1 04:05 ddf_20260501_0405.json -rw-r--r-- 1 root root 2045 May 2 04:05 ddf_20260502_0405.json -rw-r--r-- 1 root root 2046 May 3 04:05 ddf_20260503_0405.json -rw-r--r-- 1 root root 2045 May 4 04:05 ddf_20260504_0405.json -rw-r--r-- 1 root root 2046 May 5 04:05 ddf_20260505_0405.json -rw-r--r-- 1 root root 2046 May 5 06:36 ddf_20260505_0636.json -rw-r--r-- 1 root root 2046 May 6 04:05 ddf_20260506_0405.json -rw-r--r-- 1 root root 3344 Apr 17 08:13 evolution_20260417_0803.md -rw-r--r-- 1 root root 3248 Apr 18 03:39 evolution_20260418_0330.md -rw-r--r-- 1 root root 2793 Apr 19 08:07 evolution_20260419_0330.md -rw-r--r-- 1 root root 3232 Apr 20 03:35 evolution_20260420_0330.md -rw-r--r-- 1 root root 2507 Apr 21 03:37 evolution_20260421_0330.md -rw-r--r-- 1 root root 3132 Apr 22 03:36 evolution_20260422_0330.md -rw-r--r-- 1 root root 2475 Apr 22 16:25 evolution_20260422_1616.md -rw-r--r-- 1 root root 2507 Apr 23 03:35 evolution_20260423_0330.md -rw-r--r-- 1 root root 2619 Apr 24 03:47 evolution_20260424_0330.md -rw-r--r-- 1 root root 2462 Apr 25 03:39 evolution_20260425_0330.md -rw-r--r-- 1 root root 3040 Apr 27 03:37 evolution_20260427_0330.md -rw-r--r-- 1 root root 2305 Apr 28 03:40 evolution_20260428_0330.md -rw-r--r-- 1 root root 2680 May 3 03:40 evolution_20260503_0330.md -rw-r--r-- 1 root root 2406 May 4 09:08 evolution_20260504_0330.md -rw-r--r-- 1 root root 2548 May 5 03:34 evolution_20260505_0330.md -rw-r--r-- 1 root root 2996 May 6 03:37 evolution_20260506_0330.md -rw-r--r-- 1 root root 21634 Apr 8 03:34 exp_acf_decay_data.json -rw-r--r-- 1 root root 4386 Apr 5 08:27 exp_boundary_20260405_0825.json -rw-r--r-- 1 root root 1180 Apr 5 09:17 exp_boundary_growth_20260405_0914.json -rw-r--r-- 1 root root 14808 Apr 5 09:21 exp_brody_crossover_20260405.json -rw-r--r-- 1 root root 14041 Apr 6 07:16 exp_crossover_universality.json -rw-r--r-- 1 root root 10183 Apr 6 07:12 exp_dR_brody_connection.json -rw-r--r-- 1 root root 9693 Apr 6 07:09 exp_desitter_unification.json -rw-r--r-- 1 root root 11031 Apr 5 09:17 exp_excess_scaling_20260405.json -rw-r--r-- 1 root root 916 Apr 5 12:07 exp_geodesic_deviation_primes.json -rw-r--r-- 1 root root 7207 Apr 6 07:02 exp_metric_tensor_diag_long.json -rw-r--r-- 1 root root 2181 Apr 5 11:45 exp_number_variance_test.json -rw-r--r-- 1 root root 471174 Apr 8 06:47 exp_psd_prime_gaps.json -rw-r--r-- 1 root root 3804 Apr 5 11:47 exp_ricci_primes.json -rw-r--r-- 1 root root 3668 Apr 29 10:20 falsifier_20260429_1013.json -rw-r--r-- 1 root root 2846 Apr 29 10:50 falsifier_20260429_1041.json -rw-r--r-- 1 root root 2370 Apr 30 03:36 falsifier_20260430_0330.json -rw-r--r-- 1 root root 3953 Apr 30 19:11 falsifier_20260430_1905.json -rw-r--r-- 1 root root 4165 Apr 30 19:25 falsifier_20260430_1919.json -rw-r--r-- 1 root root 3965 Apr 30 19:55 falsifier_20260430_1946.json -rw-r--r-- 1 root root 3852 May 1 03:37 falsifier_20260501_0330.json -rw-r--r-- 1 root root 4323 May 1 07:41 falsifier_20260501_0725.json -rw-r--r-- 1 root root 3847 May 1 09:39 falsifier_20260501_0931.json -rw-r--r-- 1 root root 3823 May 2 03:36 falsifier_20260502_0330.json -rw-r--r-- 1 root root 4228 May 3 03:40 falsifier_20260503_0330.json -rw-r--r-- 1 root root 5793 May 4 09:08 falsifier_20260504_0901.json -rw-r--r-- 1 root root 383 May 4 12:29 falsifier_20260504_1219.json -rw-r--r-- 1 root root 2978 May 5 03:33 falsifier_20260505_0330.json -rw-r--r-- 1 root root 28 May 6 03:37 falsifier_20260506_0330.raw.txt -rw-r--r-- 1 root root 330 May 6 06:32 falsifier_20260506_0625.json -rw-r--r-- 1 root root 10050 Mar 5 07:56 fibonacci_spectrum_20260305_0756.json -rw-r--r-- 1 root root 10050 Mar 5 08:44 fibonacci_spectrum_20260305_0844.json -rw-r--r-- 1 root root 10050 Mar 6 03:42 fibonacci_spectrum_20260306_0342.json -rw-r--r-- 1 root root 10050 Mar 6 18:34 fibonacci_spectrum_20260306_1834.json -rw-r--r-- 1 root root 25252 Mar 5 08:43 gap_labeling_20260305_0843.json -rw-r--r-- 1 root root 25252 Mar 5 11:11 gap_labeling_20260305_1111.json -rw-r--r-- 1 root root 25252 Mar 6 03:41 gap_labeling_20260306_0341.json -rw-r--r-- 1 root root 25252 Mar 6 18:34 gap_labeling_20260306_1834.json -rw-r--r-- 1 root root 25252 Mar 7 03:42 gap_labeling_20260307_0342.json -rw-r--r-- 1 root root 25252 Mar 15 03:43 gap_labeling_20260315_0343.json -rw-r--r-- 1 root root 6315 Apr 7 06:40 hierarchy_data.json -rw-r--r-- 1 root root 517 May 4 07:41 incident_20260504_0721.md -rw-r--r-- 1 root root 2301 May 4 12:02 incident_20260504_1138.md -rw-r--r-- 1 root root 2938 Mar 5 08:52 insights_20260305_0852.json -rw-r--r-- 1 root root 2938 Mar 6 03:42 insights_20260306_0342.json -rw-r--r-- 1 root root 2938 Mar 6 18:34 insights_20260306_1834.json -rw-r--r-- 1 root root 2938 Mar 7 03:42 insights_20260307_0342.json -rw-r--r-- 1 root root 2938 Mar 15 03:46 insights_20260315_0346.json -rw-r--r-- 1 root root 2938 Mar 29 03:43 insights_20260329_0343.json -rw-r--r-- 1 root root 2938 Apr 1 03:46 insights_20260401_0346.json -rw-r--r-- 1 root root 2938 Apr 3 03:30 insights_20260403_0330.json -rw-r--r-- 1 root root 2938 Apr 5 07:29 insights_20260405_0729.json -rw-r--r-- 1 root root 5254 Mar 5 10:42 lagrangiana_20260305_1042.json -rw-r--r-- 1 root root 7664 Mar 5 10:48 lagrangiana_20260305_1048.json lrwxrwxrwx 1 root root 22 May 6 06:32 latest.md -> agent_20260506_0625.md -rw-r--r-- 1 root root 1257 May 6 19:41 manual_run_20260506_194105.log -rw-r--r-- 1 root root 9210 Apr 21 08:31 mapping_validation_2026-04-21.json -rw-r--r-- 1 root root 1499 Mar 5 11:11 next_exec_20260305_1111.json -rw-r--r-- 1 root root 1272 Mar 6 03:42 next_exec_20260306_0342.json -rw-r--r-- 1 root root 1446 Mar 6 18:34 next_exec_20260306_1834.json -rw-r--r-- 1 root root 1287 Mar 7 03:42 next_exec_20260307_0342.json -rw-r--r-- 1 root root 1203 Mar 14 03:42 next_exec_20260314_0342.json -rw-r--r-- 1 root root 1253 Mar 15 03:46 next_exec_20260315_0346.json -rw-r--r-- 1 root root 1168 Mar 26 03:43 next_exec_20260326_0343.json -rw-r--r-- 1 root root 1168 Mar 27 03:44 next_exec_20260327_0344.json -rw-r--r-- 1 root root 1168 Mar 28 03:44 next_exec_20260328_0344.json -rw-r--r-- 1 root root 1329 Mar 29 03:43 next_exec_20260329_0343.json -rw-r--r-- 1 root root 987 Mar 30 03:44 next_exec_20260330_0344.json -rw-r--r-- 1 root root 987 Mar 31 03:45 next_exec_20260331_0345.json -rw-r--r-- 1 root root 1362 Apr 1 03:46 next_exec_20260401_0346.json -rw-r--r-- 1 root root 1452 Apr 2 03:44 next_exec_20260402_0344.json -rw-r--r-- 1 root root 1207 Apr 3 03:30 next_exec_20260403_0330.json -rw-r--r-- 1 root root 423 Apr 4 03:30 next_exec_20260404_0330.json -rw-r--r-- 1 root root 368 Apr 5 03:30 next_exec_20260405_0330.json -rw-r--r-- 1 root root 1391 Apr 5 07:29 next_exec_20260405_0729.json -rw-r--r-- 1 root root 3805 Mar 6 09:02 phi_vs_silver_falsification_20260306.json -rw-r--r-- 1 root root 17274 May 6 06:33 recovery_20260506_062520.log -rw-r--r-- 1 root root 1806 Mar 2 03:41 report_20260302_0341.md -rw-r--r-- 1 root root 1775 Mar 3 03:41 report_20260303_0341.md -rw-r--r-- 1 root root 1791 Mar 4 03:42 report_20260304_0342.md -rw-r--r-- 1 root root 1775 Mar 5 03:42 report_20260305_0342.md -rw-r--r-- 1 root root 1791 Mar 5 21:21 report_20260305_2121.md -rw-r--r-- 1 root root 1776 Mar 6 03:41 report_20260306_0341.md -rw-r--r-- 1 root root 1791 Mar 7 03:42 report_20260307_0342.md -rw-r--r-- 1 root root 1790 Mar 14 03:42 report_20260314_0342.md -rw-r--r-- 1 root root 1789 Mar 15 03:42 report_20260315_0342.md -rw-r--r-- 1 root root 1743 Mar 15 08:01 report_20260315_0801.md -rw-r--r-- 1 root root 1765 Mar 26 03:43 report_20260326_0343.md -rw-r--r-- 1 root root 1763 Mar 27 03:44 report_20260327_0344.md -rw-r--r-- 1 root root 1779 Mar 28 03:44 report_20260328_0344.md -rw-r--r-- 1 root root 1677 Mar 29 03:43 report_20260329_0343.md -rw-r--r-- 1 root root 1514 Mar 30 03:44 report_20260330_0344.md -rw-r--r-- 1 root root 1656 Mar 31 03:45 report_20260331_0345.md -rw-r--r-- 1 root root 1671 Mar 31 18:09 report_20260331_1809.md -rw-r--r-- 1 root root 1663 Apr 1 03:46 report_20260401_0346.md -rw-r--r-- 1 root root 1656 Apr 2 03:44 report_20260402_0344.md -rw-r--r-- 1 root root 1656 Apr 2 07:56 report_20260402_0756.md -rw-r--r-- 1 root root 1656 Apr 3 03:30 report_20260403_0330.md -rw-r--r-- 1 root root 1656 Apr 4 03:30 report_20260404_0330.md -rw-r--r-- 1 root root 1656 Apr 4 18:52 report_20260404_1852.md -rw-r--r-- 1 root root 1656 Apr 5 03:30 report_20260405_0330.md -rw-r--r-- 1 root root 1656 Apr 5 07:15 report_20260405_0715.md -rw-r--r-- 1 root root 2593 Apr 21 15:11 tension_projection_screening_2026-04-21.json -rw-r--r-- 1 root root 1116 Mar 5 08:43 trace_map_20260305_0843.json -rw-r--r-- 1 root root 1186 Mar 5 08:44 trace_map_20260305_0844.json -rw-r--r-- 1 root root 458 May 6 19:31 trajectory_apply_20260506_1931.json -rw-r--r-- 1 root root 458 May 6 19:41 trajectory_apply_20260506_1941.json -rw-r--r-- 1 root root 294 May 6 04:30 watchdog.log codex Il campo punta a META: il rischio attivo non è “trovare un altro segnale”, ma distinguere segnale stabile da artefatto di definizione/campionamento. Ora eseguo la proiezione e controllo gli script/osservabili canonici prima di tagliare l’esperimento. exec /bin/bash -lc "sed -n '1,260p' tools/observables_registry.py" in /opt/MM_D-ND succeeded in 0ms: """observables_registry.py — Source of Truth per gli observables del lab D-ND. Cristallizzato 2026-05-06 dalla **consecutio del cycle agent_20260506_0625**: > "What opens now: the lab needs an observable registry. Labels like SR > cannot travel between reports unless they bind to a function definition. > Without that, META flags are not philosophical: the same label can > silently change the object under test." ## Il problema che ha creato il registry Il cycle 06:25 ha auto-falsificato il finding del cycle 03:30 ("secondo asse GUE") e nel farlo ha trovato **collision di nomi observable** tra script: - `SR` in `exp_selective_layer_decoupling.py` = `spacing_ratio` (mean min/max ratio of consecutive gaps) — convention dominante (~6 script) - `SR` in `exp_scale_selective_perturbation.py` = `spectral_rigidity(gaps)` (Δ₃(L) rigidity) — variante usata SOLO in 1 script - `triple_var` in 3 script = `np.var(triple_sums)` (raw) — convention dominante - `triple_var` in `exp_perturbation_dimensionality_audit.py` = `np.var(triples) / np.var(gaps)` (normalizzato) — variante in 1 script Il lab autonomo che compara report tra script con osservabili "stesso nome, funzione diversa" stava confrontando mele con arance. ## La soluzione (minimal, non invasiva) Questo registry stabilisce il **nome canonico**: ciò che la maggioranza degli script chiama già `SR`/`triple_var`/etc. Le varianti restano disponibili ma con nomi ESPLICITI (`SR_local_rigidity`, `triple_var_normalized`) per evitare mascheramento semantico. ## Come usarlo ```python from observables_registry import OBSERVABLES_CANONICAL, OBSERVABLES_REGISTRY_VERSION # Compute canonical observable suite for a sequence of gaps results = {name: fn(gaps) for name, fn in OBSERVABLES_CANONICAL.items()} # Or import individual canonical observable from observables_registry import SR, triple_var, L1, L2, SR2 # For variants, import explicitly with disambiguating name from observables_registry import SR_local_rigidity, triple_var_normalized ``` ## Convention per i report Ogni report agent (cycle) che usa observables DEVE includere nel suo header: ``` observables_registry: 1.0.0-2026-05-06 observables_used: [SR, SR2, L1, L2, triple_var] ``` Cycle che mescola canonical + variant DEVE indicare entrambi: ``` observables_used: [SR, SR_local_rigidity, ...] ``` Senza questo, i confronti cross-cycle sono inattendibili. ## Versioning Cambiare una definizione canonica = bump del registry version e nota nel changelog. Le definizioni canoniche sono **immutabili dentro una versione**. """ from __future__ import annotations import numpy as np OBSERVABLES_REGISTRY_VERSION = "1.0.0-2026-05-06" # ─── Canonical observables (convention dominante nel codebase 2026-05-06) ─── def SR(gaps: np.ndarray) -> float: """**SR — Spacing Ratio** (canonical). Mean of `min(g_i, g_{i+1}) / max(g_i, g_{i+1})` over consecutive gaps. Range: (0, 1]. GUE → ~0.60. Poisson → ~0.39. Picket-fence → 1. NOTE: questa è la convention dominante in 6+ script del lab. Per la variante "local spectral rigidity Δ₃(L)" usare `SR_local_rigidity`. """ if len(gaps) < 2: return 0.0 s, s1 = gaps[:-1], gaps[1:] r = np.minimum(s, s1) / np.maximum(s, s1) r = r[np.isfinite(r) & (r > 0)] return float(np.mean(r)) if len(r) else 0.0 def SR2(gaps: np.ndarray) -> float: """**SR2 — Next-nearest Spacing Ratio** (canonical). Mean of `min(g_i, g_{i+2}) / max(g_i, g_{i+2})` skipping one gap. Probes lag-2 spacing structure. """ if len(gaps) < 3: return 0.0 s, s2 = gaps[:-2], gaps[2:] r = np.minimum(s, s2) / np.maximum(s, s2) r = r[np.isfinite(r) & (r > 0)] return float(np.mean(r)) if len(r) else 0.0 def L1(gaps: np.ndarray) -> float: """**L1 — Lag-1 Autocorrelation** (canonical). Standard ACF at lag 1 of the gap sequence. """ if len(gaps) < 3: return 0.0 g = gaps - np.mean(gaps) c0 = float(np.mean(g ** 2)) if c0 <= 1e-15: return 0.0 return float(np.mean(g[:-1] * g[1:]) / c0) def L2(gaps: np.ndarray) -> float: """**L2 — Lag-2 Autocorrelation** (canonical).""" if len(gaps) < 4: return 0.0 g = gaps - np.mean(gaps) c0 = float(np.mean(g ** 2)) if c0 <= 1e-15: return 0.0 return float(np.mean(g[:-2] * g[2:]) / c0) def triple_var(gaps: np.ndarray) -> float: """**triple_var — Variance of consecutive gap triples** (canonical). Variance of `g_i + g_{i+1} + g_{i+2}` over the sequence (RAW, no normalization). Convention used in 3+ scripts. For the normalized version (variance ratio `var(triples) / var(gaps)`) use `triple_var_normalized`. """ if len(gaps) < 3: return 0.0 t = gaps[:-2] + gaps[1:-1] + gaps[2:] return float(np.var(t)) # Set canonico per uso "compute all" da report OBSERVABLES_CANONICAL: dict[str, callable] = { "SR": SR, "SR2": SR2, "L1": L1, "L2": L2, "triple_var": triple_var, } # ─── Variants (esplicitamente nominate, no collision con canonical) ─── def SR_local_rigidity(gaps: np.ndarray, L: int = 10) -> float: """**SR_local_rigidity — Δ₃(L) Spectral Rigidity** (variant). Different observable than canonical `SR` (spacing ratio). Measures the average squared deviation of the cumulative spacing function from the best-fit straight line in a window of size L. Originated from `exp_scale_selective_perturbation.py` where it was locally named `SR` — registered here as `SR_local_rigidity` to avoid collision with canonical spacing-ratio definition. Use when explicitly studying spectral rigidity, NOT as alias for SR. """ if len(gaps) < 5: return 0.0 cumulative = np.cumsum(gaps) if cumulative[-1] <= 1e-15: return 0.0 cumulative = cumulative / cumulative[-1] * len(cumulative) n = np.arange(1, len(cumulative) + 1, dtype=float) window = int(min(L * len(gaps) / cumulative[-1], len(gaps) // 2)) if window < 5: return 0.0 residuals = [] for start in range(0, len(cumulative) - window, max(1, window // 2)): end = start + window x = n[start:end] y = cumulative[start:end] a, b = np.polyfit(x, y, 1) residuals.append(np.mean((y - (a * x + b)) ** 2)) return float(np.mean(residuals)) if residuals else 0.0 def triple_var_normalized(gaps: np.ndarray) -> float: """**triple_var_normalized — Variance of triples / variance of gaps** (variant). Originated from `exp_perturbation_dimensionality_audit.py` where it was locally named `triple_var` — registered here as `triple_var_normalized` to avoid collision with canonical raw triple variance. Use when explicitly studying triple-variance scaling relative to single-gap variance, NOT as alias for triple_var. """ if len(gaps) < 3: return 0.0 triples = gaps[:-2] + gaps[1:-1] + gaps[2:] v = float(np.var(gaps)) if v <= 1e-15: return 0.0 return float(np.var(triples) / v) # Set varianti, importabile esplicitamente OBSERVABLES_VARIANTS: dict[str, callable] = { "SR_local_rigidity": SR_local_rigidity, "triple_var_normalized": triple_var_normalized, } # ─── Public API ─────────────────────────────────────────────────────── def compute_canonical(gaps: np.ndarray) -> dict[str, float]: """Compute all canonical observables for a gap sequence. Returns dict {name: value} ready for inclusion in cycle reports. """ return {name: fn(gaps) for name, fn in OBSERVABLES_CANONICAL.items()} def report_header() -> str: """Suggested markdown header line for cycle reports using this registry.""" canonical_list = ", ".join(OBSERVABLES_CANONICAL.keys()) return ( f"observables_registry: {OBSERVABLES_REGISTRY_VERSION}\n" f"observables_used: [{canonical_list}]" ) __all__ = [ "OBSERVABLES_REGISTRY_VERSION", "OBSERVABLES_CANONICAL", "OBSERVABLES_VARIANTS", "SR", "SR2", "L1", "L2", "triple_var", "SR_local_rigidity", "triple_var_normalized", "compute_canonical", "report_header", ] if __name__ == "__main__": # Smoke test: canonical observables on a simple gap series rng = np.random.default_rng(42) gue_like = rng.gamma(shape=2.0, scale=0.5, size=200) exec /bin/bash -lc "sed -n '1,220p' tools/data/reports/agent_20260506_0330.md" in /opt/MM_D-ND succeeded in 0ms: # Agent Report — Scale-Selective Perturbations Reveal a Second Axis in GUE, Not in Primes **Date**: 2026-05-06 03:30 **Piano**: 65 **Tension explored**: META (0.5) + BOUNDARY (0.8) ## Claim Under Test > The single latent coordinate found under uniform partial shuffle (PC1 ~99%, observable rank audit 05-05) is an artifact of uniform shuffle collapsing multiple perturbation axes into one. Scale-selective perturbations should reveal multi-dimensional structure. ## Question Do structurally different perturbations (adjacent-swap, block-shuffle, large-gap-only, uniform) produce different observable profiles, or do they all collapse to the same retention curve? ## Experiment Design - **Four perturbation types**: 1. Adjacent-swap: swap neighboring pairs with probability alpha. Destroys lag-1 correlations selectively, preserves long-range order. 2. Block-shuffle: shuffle within blocks of 50 elements. Alpha controls fraction of blocks shuffled. Destroys intra-block order, preserves inter-block structure. 3. Large-gap-only: shuffle positions of above-median gaps only. Preserves small-gap ordering entirely. 4. Uniform partial: baseline from previous runs. - **Alphas**: 0.1, 0.3, 0.5, 0.7, 0.9; 16 trials per (perturbation, alpha) pair. - **Observables**: SR, L1, L2, SR2, triple_var — same set as observable rank audit. - **Null baseline**: 48 full shuffles per domain. - **Domains**: 30,000 prime gaps; GUE (253 unfolded spacings from 23x23 Hermitian matrix). - **PCA**: across all 20 profiles (4 perturbation types x 5 alphas), retention-normalized. Effective rank = exp(entropy of singular value distribution). - **Centroid cosine similarity**: mean retention vector per perturbation type, pairwise cosine. - **Seed**: 20260506. ## Results ### Primes (N=30,000) | Perturbation | SR | L1 | L2 | SR2 | triple_var | |---|---:|---:|---:|---:|---:| | adjacent_swap | 1.039 | 0.886 | 1.399 | 1.003 | 1.246 | | block_shuffle | 0.581 | 0.404 | 0.298 | 0.514 | 0.342 | | large_gap_only | 0.890 | 0.883 | 0.892 | 0.826 | 0.988 | | uniform | 0.307 | 0.376 | 0.376 | 0.358 | 0.237 | *Retention at alpha=0.5. Values >1 mean the perturbation enhanced the signal relative to unperturbed.* PCA: PC1 = 95.4%, PC2 = 2.6%, PC3 = 1.3%. Effective rank = 1.263. All centroid cosine similarities > 0.955. ### GUE (N=253) | Perturbation | SR | L1 | L2 | SR2 | triple_var | |---|---:|---:|---:|---:|---:| | adjacent_swap | 0.918 | 0.784 | -1.135 | 0.941 | 0.985 | | block_shuffle | 0.804 | 0.693 | 0.455 | 0.778 | 0.832 | | large_gap_only | 0.671 | 0.859 | 1.173 | 0.712 | 0.800 | | uniform | 0.449 | 0.423 | 0.634 | 0.353 | 0.435 | *Retention at alpha=0.5.* PCA: PC1 = 73.5%, PC2 = 25.2%, PC3 = 1.1%. Effective rank = 1.889. Adjacent_swap vs large_gap_only cosine = 0.667 (substantially non-aligned). ## Key Findings 1. **GUE has a second perturbation axis that primes do not.** Under scale-selective perturbations, GUE effective rank rises from ~1.02 (uniform-only, rank audit 05-05) to 1.889. The second principal component explains 25.2% of variance. For primes, the rank rises only from 1.07 to 1.26 — the second axis is weak (2.6%). 2. **The L2 observable discriminates perturbation types in GUE.** Under adjacent-swap, GUE L2 retention = -1.135 (sign flip: the perturbation reverses lag-2 correlation). Under large-gap-only, L2 retention = +1.173 (enhancement). This sign difference means adjacent-swap and large-gap-only probe structurally distinct axes of GUE correlations. For primes, L2 also shows anomalous behavior (1.399 under adjacent-swap) but does not flip sign. 3. **Adjacent-swap is the most selective perturbation.** In primes, adjacent swapping barely touches SR (retention 1.039) and SR2 (1.003) while reducing L1 to 0.886 and enhancing L2 to 1.399. The enhancement of L2 under adjacent-swap is not trivial: swapping neighbors creates new lag-2 correlations from the original lag-1 structure (if g_n,g_{n+1} swap, the new g_{n+1} becomes the old g_n, creating a new lag-2 pair from the old lag-1 pair). 4. **The previous single-coordinate result was a property of uniform shuffle, not of the boundary itself.** Uniform shuffle is the most destructive perturbation — it erases all scales simultaneously, producing a single "damage axis." Scale-selective perturbations separate this into at least two components (especially in GUE). 5. **Caveat: GUE sample is small (N=253).** The GUE matrix size was limited by available computation. The effective rank 1.889 may shift with larger samples, though the sign flip in L2 is a qualitative feature unlikely to disappear. ## Verdict **CONSTRAINT on META + BOUNDARY**: the single latent coordinate found under uniform shuffle (rank audit 05-05) is a property of the perturbation type, not of the boundary itself. Scale-selective perturbations reveal a second axis in GUE (PC2=25.2%) and a weak second axis in primes (PC2=2.6%). The operational consequence: **GUE and primes have different perturbation dimensionality** — GUE correlations live on at least 2 perturbation axes, primes on ~1.3. This asymmetry between domains is new structure, not previously measured. Perimeter: 30,000 prime gaps (p_2 to p_{30001}), 253 GUE spacings, 4 perturbation types, 5 alpha values, 16 trials each, seed 20260506. ## Bicono della scoperta - **Due radici**: uniform shuffle (one axis, all scales destroyed) · scale-selective perturbation (multiple axes, each scale targeted separately). The duality is: homogeneous destruction vs structured probing. - **Singolare**: the correlation structure of the sequence before any perturbation is applied. It contains all axes simultaneously — the perturbation type selects which to reveal. - **Invariante di passaggio**: GUE has higher perturbation dimensionality than primes. This holds across the perturbation types tested. The asymmetry (GUE ~2D, primes ~1.3D) persists through the vertex. - **Campo di possibilita**: here it becomes possible to distinguish domains by HOW they respond to structured probing (not just WHETHER they respond). Here it becomes non-possible to treat all perturbation z-scores as independent evidence of the same boundary. ## Auto-audit: 5 lenti - **L1 hard constraint vs bias**: No absolutes. "~2D" and "~1.3D" are effective ranks, not exact integers. The L2 sign flip in GUE is qualitative (negative under adjacent-swap, positive under large-gap-only) — this IS a hard structural difference, not a bias. - **L2 quantita vs ratio**: Retentions are ratios normalized by (original - shuffle_mean). The denominators differ across observables but are used consistently within each observable. No cross-space ratio comparison. - **L3 no silent patching**: The rank audit claim (05-05: "boundary is 1D in observable space") is NOT declared resolved. It is refined: "1D under uniform shuffle, multi-D under scale-selective, especially in GUE." The distinction is explicit. - **L4 edge cases**: GUE N=253 is declared as caveat. The sign flip in L2 is not an edge case — it occurs at all tested alphas (0.1 through 0.9 for adjacent-swap). - **L5 re-discovery**: Scale-selective perturbation is related to multiscale analysis (wavelet-like decompositions). The specific application to prime gap vs GUE spacing perturbation dimensionality is not a standard result in random matrix theory. The L2 sign flip under adjacent-swap may relate to Markov chain perturbation theory. Not tagged NEW — tagged as CONSTRAINT refining previous result. ## Files - Script: `tools/exp_scale_selective_perturbation.py` - Data: `tools/data/scale_selective_perturbation.json` - Report: `tools/data/reports/agent_20260506_0330.md` exec /bin/bash -lc "sed -n '1,220p' tools/data/reports/agent_20260506_0625.md" in /opt/MM_D-ND succeeded in 0ms: # Agent Report — Perturbation Dimensionality Is Not Yet a Stable GUE Invariant **Date**: 2026-05-06 06:25 **Piano**: 65 **Tension explored**: META + BOUNDARY ## Claim Under Test The 03:30 run reported that scale-selective perturbations reveal a second axis in GUE (effective rank 1.889, PC2=25.2%) but not in primes. The caveat was explicit: the GUE sample was short (`N=253`). This run tests whether that second axis survives independent GUE ensembles and whether it depends on the observable definitions hidden under the same labels (`SR`, `SR2`, `triple_var`). ## Experiment Tool created: `tools/exp_perturbation_dimensionality_audit.py` Perimeter atomico: - fixed domains: `primes` 12,000 gaps, `prime_shuffle_control` 12,000 permuted prime gaps, `poisson` 12,000 iid exponential spacings; - GUE long control: 6 independent replicates, each from 16 Hermitian matrices of size 180, after 10% edge trim; each replicate has 2,288 spacings; - GUE short control: 6 independent one-matrix replicates of size 42, after edge trim; this approximates the fragile small-sample regime; - perturbations: `adjacent_swap`, `block_shuffle`, `large_gap_only`, `uniform`; - alpha grid: 0.1, 0.3, 0.5, 0.7, 0.9; - trials per perturbation-alpha: 10; full-shuffle baseline: 24; - seed: 20260506. Two observable sets were run because a META issue emerged: - `rank_audit`: `SR` = nearest-neighbor spacing ratio, `SR2` = next-nearest spacing ratio, `triple_var` = normalized variance of triple sums. - `scale_0330`: `SR` = local spectral rigidity at L=10, `SR2` = local spectral rigidity at L=20, `triple_var` = variance of triple products. ## Results ### Rank-audit observable set | Domain | N | Effective rank | PC2 | adjacent vs large cosine | |---|---:|---:|---:|---:| | primes | 12000 | 1.374 | 0.070 | 0.947 | | prime_shuffle_control | 12000 | 2.294 | 0.199 | 0.247 | | poisson | 12000 | 1.917 | 0.193 | 0.918 | | GUE long, 6 reps mean | 2288 each | 1.305 ± 0.278 | 0.064 ± 0.066 | 0.877 ± 0.081 | | GUE short, 6 reps mean | short | 1.683 ± 0.498 | 0.106 ± 0.080 | 0.567 ± 0.340 | ### Scale-0330 observable set | Domain | N | Effective rank | PC2 | adjacent vs large cosine | |---|---:|---:|---:|---:| | primes | 12000 | 1.318 | 0.046 | 0.975 | | prime_shuffle_control | 12000 | 1.988 | 0.085 | 0.526 | | poisson | 12000 | 2.201 | 0.198 | 0.885 | | GUE long, 6 reps mean | 2288 each | 1.381 ± 0.223 | 0.099 ± 0.069 | 0.874 ± 0.082 | | GUE short, 6 reps mean | short | 2.013 ± 0.525 | 0.159 ± 0.087 | 0.746 ± 0.242 | ## Findings 1. **The strong GUE second-axis claim does not survive as stated.** Under direct `scale_0330` observables, long independent GUE replicates give rank 1.381 ± 0.223 and PC2 9.9% ± 6.9%, not rank 1.889 and PC2 25.2%. The previous number is inside the fragile short-sample regime: GUE short controls have rank 2.013 ± 0.525 and PC2 15.9% ± 8.7%. 2. **Short GUE samples inflate apparent perturbation dimensionality.** In both observable sets, GUE short has higher rank and larger variance than GUE long. This does not prove the 03:30 axis was false in every configuration; it restricts it to a sample-size-sensitive observation unless a larger-replicate run recovers it. 3. **The lab has an observable-name collision.** `SR`, `SR2`, and `triple_var` do not name the same functions across the recent scripts. `exp_observable_rank_audit.py` uses spacing-ratio and triple-sum variance; `exp_scale_selective_perturbation.py` uses local spectral rigidity and triple-product variance. Therefore the sentence "same observables as observable rank audit" in the 03:30 report is not exact. This is a META constraint, not a numerical subtlety. 4. **Primes remain close to one perturbation coordinate in both observable sets.** Primes rank is 1.374 in `rank_audit` and 1.318 in `scale_0330`; PC2 is 7.0% and 4.6%. This part of the 03:30 asymmetry is stable in the tested perimeter. 5. **Poisson and shuffled-prime controls show multi-axis artifacts.** Poisson has rank 1.917/2.201 depending on observable set; prime shuffle control has rank 2.294/1.988. Multi-dimensional perturbation response by itself is not evidence of structured GUE-like boundary. It can arise from low structural signal plus noisy denominators in retention normalization. ## Verdict **CONSTRAINT on META + BOUNDARY**: "GUE has a second perturbation axis" must be scoped to the exact sample length, generator, and observable definitions. In the larger independent-GUE perimeter tested here, the robust statement is weaker: > Primes remain near one perturbation coordinate under both observable sets; GUE long replicates show only a weak second component; short GUE samples can inflate apparent rank; Poisson and shuffled controls can also appear multi-axis. The boundary is still operator-dependent, but perturbation dimensionality is not yet a stable domain invariant. The next valid test is not another single GUE matrix; it is a replicate-and-size curve for effective rank vs number of spacings, with observable definitions versioned. ## Consecutio What opens now: the lab needs an observable registry. Labels like `SR` cannot travel between reports unless they bind to a function definition. Without that, META flags are not philosophical: the same label can silently change the object under test. ## Auto-audit: 5 lenti - **L1 hard constraint vs bias**: no zero/always claim. "Does not survive as stated" means the reported magnitude did not reproduce in the stated larger perimeter. - **L2 quantity vs ratio**: effective rank and PC2 are reported with sample size and replicate variance. Retention ratios are not treated as absolute structure without controls. - **L3 no silent patching**: the 03:30 claim is explicitly restricted, not renamed as confirmed. The observable-name mismatch is declared. - **L4 edge cases**: the short-GUE effect is isolated as its own control, not averaged into long GUE. - **L5 re-discovery**: PCA/effective-rank instability under small samples is a known statistical issue. This report is a lab constraint on framing, not a NEW mathematical result. ## Files - Script: `tools/exp_perturbation_dimensionality_audit.py` - Data: `tools/data/perturbation_dimensionality_audit.json` - Data: `tools/data/perturbation_dimensionality_audit_scale0330.json` - Report: `tools/data/reports/agent_20260506_0625.md` exec /bin/bash -lc 'python tools/dnd_scenario.py --best' in /opt/MM_D-ND succeeded in 150ms: META: score=0.898 — META: incerto (i=0.5) — massimo potere discriminante G_POTENZIALE_NULLA: score=0.703 — G_POTENZIALE_NULLA: legge di scala da raffinare METRIC_TENSOR: score=0.624 — METRIC_TENSOR: legge di scala da raffinare codex La consecutio è chiara: non serve un altro confronto singolo, serve una curva dimensione-campione. Creo uno strumento riusabile che misura rank/PC2 contro N usando osservabili canonici versionati, con GUE, Poisson, primi e shuffle-primi nello stesso perimetro. exec /bin/bash -lc "sed -n '1,280p' tools/exp_perturbation_dimensionality_audit.py" in /opt/MM_D-ND succeeded in 0ms: #!/usr/bin/env python3 """ exp_perturbation_dimensionality_audit.py Robustness audit for the scale-selective perturbation result. The 2026-05-06 03:30 run found that GUE spacing sequences expose a second perturbation axis under scale-selective probes, while prime gaps remain close to one axis. That run used a short GUE sequence. This tool repeats the same kind of measurement across independent replicates and explicit sample-size controls. It measures only observables and null baselines. The report owns the claim. """ from __future__ import annotations import argparse import json from pathlib import Path import numpy as np OBS_NAMES = ["SR", "L1", "L2", "SR2", "triple_var"] PERT_NAMES = ["adjacent_swap", "block_shuffle", "large_gap_only", "uniform"] OBSERVABLE_SET = "rank_audit" def prime_gaps(n_gaps: int) -> np.ndarray: limit = max(100, int(n_gaps * (np.log(n_gaps + 10) + np.log(np.log(n_gaps + 10)) + 5))) while True: sieve = np.ones(limit + 1, dtype=bool) sieve[:2] = False for p in range(2, int(limit**0.5) + 1): if sieve[p]: sieve[p * p : limit + 1 : p] = False primes = np.flatnonzero(sieve) if len(primes) >= n_gaps + 1: return np.diff(primes[: n_gaps + 1]).astype(float) limit *= 2 def gue_spacings(matrix_size: int, n_matrices: int, rng: np.random.Generator) -> np.ndarray: parts = [] edge = max(2, matrix_size // 10) for _ in range(n_matrices): real = rng.standard_normal((matrix_size, matrix_size)) imag = rng.standard_normal((matrix_size, matrix_size)) h = real + 1j * imag h = (h + h.conj().T) / (2.0 * np.sqrt(matrix_size)) eigs = np.sort(np.linalg.eigvalsh(h).real) bulk = eigs[edge:-edge] gaps = np.diff(bulk) mean = np.mean(gaps) if mean > 1e-15: parts.append(gaps / mean) return np.concatenate(parts).astype(float) def spacing_ratio(gaps: np.ndarray, lag: int) -> float: a = gaps[:-lag] b = gaps[lag:] denom = np.maximum(a, b) valid = denom > 1e-15 if not np.any(valid): return 0.0 return float(np.mean(np.minimum(a[valid], b[valid]) / denom[valid])) def lag_acf(gaps: np.ndarray, lag: int) -> float: g = gaps - np.mean(gaps) c0 = np.dot(g, g) if c0 <= 1e-15 or len(gaps) <= lag: return 0.0 return float(np.dot(g[:-lag], g[lag:]) / c0) def triple_var(gaps: np.ndarray) -> float: if len(gaps) < 3: return 0.0 triples = gaps[:-2] + gaps[1:-1] + gaps[2:] v = np.var(gaps) if v <= 1e-15: return 0.0 return float(np.var(triples) / v) def spectral_rigidity(gaps: np.ndarray, L: int) -> float: cumulative = np.cumsum(gaps) if cumulative[-1] <= 1e-15: return 0.0 cumulative = cumulative / cumulative[-1] * len(cumulative) n = np.arange(1, len(cumulative) + 1, dtype=float) window = int(min(L * len(gaps) / cumulative[-1], len(gaps) // 2)) if window < 5: return 0.0 step = max(1, window // 2) residuals = [] for start in range(0, len(gaps) - window, step): seg_n = n[start : start + window] seg_c = cumulative[start : start + window] coeffs = np.polyfit(seg_n, seg_c, 1) fitted = np.polyval(coeffs, seg_n) residuals.append(np.mean((seg_c - fitted) ** 2)) return float(np.mean(residuals)) if residuals else 0.0 def triple_product_var(gaps: np.ndarray) -> float: if len(gaps) < 3: return 0.0 triples = gaps[:-2] * gaps[1:-1] * gaps[2:] return float(np.var(triples)) def measure(gaps: np.ndarray) -> dict[str, float]: if OBSERVABLE_SET == "scale_0330": return { "SR": spectral_rigidity(gaps, 10), "L1": lag_acf(gaps, 1), "L2": lag_acf(gaps, 2), "SR2": spectral_rigidity(gaps, 20), "triple_var": triple_product_var(gaps), } return { "SR": spacing_ratio(gaps, 1), "L1": lag_acf(gaps, 1), "L2": lag_acf(gaps, 2), "SR2": spacing_ratio(gaps, 2), "triple_var": triple_var(gaps), } def perturb_adjacent_swap(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: out = gaps.copy() idx = np.arange(0, len(out) - 1, 2) chosen = idx[rng.random(len(idx)) < alpha] tmp = out[chosen].copy() out[chosen] = out[chosen + 1] out[chosen + 1] = tmp return out def perturb_block_shuffle(gaps: np.ndarray, alpha: float, rng: np.random.Generator, block_size: int = 64) -> np.ndarray: out = gaps.copy() n_blocks = len(out) // block_size if n_blocks <= 0: return rng.permutation(out) if alpha > 0 else out k = int(round(alpha * n_blocks)) if k <= 0: return out for block in rng.choice(n_blocks, size=min(k, n_blocks), replace=False): start = block * block_size end = min(start + block_size, len(out)) rng.shuffle(out[start:end]) return out def perturb_large_gap_only(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: out = gaps.copy() idx = np.flatnonzero(out > np.median(out)) k = int(round(alpha * len(idx))) if k < 2: return out chosen = rng.choice(idx, size=min(k, len(idx)), replace=False) vals = out[chosen].copy() rng.shuffle(vals) out[chosen] = vals return out def perturb_uniform(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: out = gaps.copy() k = int(round(alpha * len(out))) if k < 2: return out chosen = rng.choice(len(out), size=min(k, len(out)), replace=False) vals = out[chosen].copy() rng.shuffle(vals) out[chosen] = vals return out PERTURB = { "adjacent_swap": perturb_adjacent_swap, "block_shuffle": perturb_block_shuffle, "large_gap_only": perturb_large_gap_only, "uniform": perturb_uniform, } def pca_summary(rows: list[dict]) -> dict: matrix = np.array([row["retention_vector"] for row in rows], dtype=float) matrix = matrix - np.mean(matrix, axis=0, keepdims=True) _, singular, vt = np.linalg.svd(matrix, full_matrices=False) energy = singular * singular if np.sum(energy) <= 1e-15: explained = np.zeros_like(energy) eff_rank = 0.0 else: explained = energy / np.sum(energy) pos = explained[explained > 1e-15] eff_rank = float(np.exp(-np.sum(pos * np.log(pos)))) centroids = {} for name in PERT_NAMES: vals = np.array([row["retention_vector"] for row in rows if row["perturbation"] == name]) centroids[name] = np.mean(vals, axis=0) cosine = {} for i, a_name in enumerate(PERT_NAMES): for b_name in PERT_NAMES[i + 1 :]: a = centroids[a_name] b = centroids[b_name] denom = np.linalg.norm(a) * np.linalg.norm(b) cosine[f"{a_name}_vs_{b_name}"] = float(np.dot(a, b) / denom) if denom > 1e-15 else 0.0 return { "explained_variance": [float(x) for x in explained], "effective_rank": eff_rank, "pc1_loadings": {name: float(vt[0, i]) for i, name in enumerate(OBS_NAMES)} if len(vt) else {}, "pc2_loadings": {name: float(vt[1, i]) for i, name in enumerate(OBS_NAMES)} if len(vt) > 1 else {}, "centroid_cosine": cosine, } def analyze(name: str, gaps: np.ndarray, alphas: list[float], n_trials: int, n_baseline: int, rng: np.random.Generator) -> dict: original = measure(gaps) baseline_vals = {obs: [] for obs in OBS_NAMES} for _ in range(n_baseline): row = measure(rng.permutation(gaps)) for obs in OBS_NAMES: baseline_vals[obs].append(row[obs]) baseline = { obs: { "mean": float(np.mean(vals)), "std": float(np.std(vals, ddof=1)) if len(vals) > 1 else 0.0, } for obs, vals in baseline_vals.items() } rows = [] for pert_name in PERT_NAMES: for alpha in alphas: vals = {obs: [] for obs in OBS_NAMES} for _ in range(n_trials): row = measure(PERTURB[pert_name](gaps, alpha, rng)) for obs in OBS_NAMES: vals[obs].append(row[obs]) means = {obs: float(np.mean(vals[obs])) for obs in OBS_NAMES} retention = {} for obs in OBS_NAMES: denom = original[obs] - baseline[obs]["mean"] retention[obs] = float((means[obs] - baseline[obs]["mean"]) / denom) if abs(denom) > 1e-12 else 0.0 rows.append( { "perturbation": pert_name, "alpha": float(alpha), "mean": means, "retention": retention, "retention_vector": [retention[obs] for obs in OBS_NAMES], } ) z = {} for obs in OBS_NAMES: std = baseline[obs]["std"] z[obs] = float((original[obs] - baseline[obs]["mean"]) / std) if std > 1e-12 else 0.0 return { "name": name, "n_gaps": int(len(gaps)), "original": original, "full_shuffle_baseline": baseline, "original_vs_shuffle_z": z, "profiles": rows, "pca": pca_summary(rows), } exec /bin/bash -lc "sed -n '1,260p' tools/exp_scale_selective_perturbation.py" in /opt/MM_D-ND succeeded in 0ms: #!/usr/bin/env python3 """ exp_scale_selective_perturbation.py — Scale-Selective Perturbation Audit Consecutio from observable_rank_audit: under uniform partial shuffle, all observables (SR, L1, L2, SR2, triple_var) collapse to 1 latent coordinate (PC1 ~99%). Is this because the boundary is genuinely 1D, or because uniform shuffle destroys all correlations along the same axis? This experiment applies FOUR structurally different perturbations: 1. Adjacent-swap: swap only neighboring pairs (destroys lag-1, preserves long-range) 2. Block-shuffle: shuffle within blocks of size B (preserves inter-block, destroys intra-block) 3. Large-gap-only: shuffle only the positions of above-median gaps (preserves small-gap ordering) 4. Uniform partial: the baseline from previous runs (for comparison) For each perturbation at multiple intensities, measure SR/L1/L2/SR2/triple_var, compute PCA across perturbation types, and check if the rank is > 1. If different perturbation types produce different observable profiles -> the boundary is multi-dimensional and uniform shuffle was collapsing it. If all perturbation types produce the same profile -> the boundary is 1D in observable space. Usage: python tools/exp_scale_selective_perturbation.py [--N 30000] [--seed 20260506] """ import argparse import json import numpy as np from pathlib import Path from sympy import nextprime def generate_primes(N): """Generate first N prime gaps.""" gaps = [] p = 2 for _ in range(N + 1): p_next = nextprime(p) gaps.append(p_next - p) p = p_next if len(gaps) >= N: break return np.array(gaps, dtype=float) def generate_gue(N, rng): """Generate N GUE gaps (eigenvalue spacings of random Hermitian matrix).""" dim = int(np.sqrt(2 * N)) + 10 H = rng.standard_normal((dim, dim)) + 1j * rng.standard_normal((dim, dim)) H = (H + H.conj().T) / 2 evals = np.sort(np.linalg.eigvalsh(H)) spacings = np.diff(evals) spacings = spacings[spacings > 0] # Unfold: normalize by local mean from scipy.ndimage import uniform_filter1d local_mean = uniform_filter1d(spacings.astype(float), size=max(50, len(spacings)//20)) local_mean[local_mean < 1e-12] = 1e-12 unfolded = spacings / local_mean if len(unfolded) >= N: return unfolded[:N] return unfolded def generate_poisson(N, rng): """Generate N Poisson (iid exponential) gaps.""" return rng.exponential(1.0, size=N) # ---------- Observables ---------- def spectral_rigidity(gaps, L=10): """Delta_3(L) spectral rigidity.""" cumulative = np.cumsum(gaps) cumulative = cumulative / cumulative[-1] * len(cumulative) n = np.arange(1, len(cumulative) + 1, dtype=float) window = int(min(L * len(gaps) / cumulative[-1], len(gaps) // 2)) if window < 5: return 0.0 residuals = [] for start in range(0, len(gaps) - window, max(1, window // 2)): seg_n = n[start:start+window] seg_c = cumulative[start:start+window] coeffs = np.polyfit(seg_n, seg_c, 1) fitted = np.polyval(coeffs, seg_n) residuals.append(np.mean((seg_c - fitted)**2)) return float(np.mean(residuals)) def lag1_autocorr(gaps): """Lag-1 autocorrelation.""" g = gaps - np.mean(gaps) c0 = np.dot(g, g) if c0 < 1e-15: return 0.0 return float(np.dot(g[:-1], g[1:]) / c0) def lag2_autocorr(gaps): """Lag-2 autocorrelation.""" g = gaps - np.mean(gaps) c0 = np.dot(g, g) if c0 < 1e-15: return 0.0 return float(np.dot(g[:-2], g[2:]) / c0) def sr2(gaps, L=20): """Spectral rigidity at larger L.""" return spectral_rigidity(gaps, L=L) def triple_variance(gaps): """Variance of triple products g_n * g_{n+1} * g_{n+2}.""" triples = gaps[:-2] * gaps[1:-1] * gaps[2:] return float(np.var(triples)) def compute_observables(gaps): """Compute all 5 observables.""" return { 'SR': spectral_rigidity(gaps), 'L1': lag1_autocorr(gaps), 'L2': lag2_autocorr(gaps), 'SR2': sr2(gaps), 'triple_var': triple_variance(gaps), } # ---------- Perturbation types ---------- def perturb_adjacent_swap(gaps, alpha, rng): """Swap adjacent pairs with probability alpha. Destroys lag-1, preserves long-range.""" g = gaps.copy() n = len(g) for i in range(0, n - 1, 2): if rng.random() < alpha: g[i], g[i+1] = g[i+1], g[i] return g def perturb_block_shuffle(gaps, alpha, rng, block_size=50): """Shuffle within blocks of size B. Alpha controls fraction of blocks shuffled.""" g = gaps.copy() n = len(g) n_blocks = n // block_size blocks_to_shuffle = rng.choice(n_blocks, size=max(1, int(alpha * n_blocks)), replace=False) for b in blocks_to_shuffle: start = b * block_size end = min(start + block_size, n) rng.shuffle(g[start:end]) return g def perturb_large_gap_only(gaps, alpha, rng): """Shuffle only above-median gap positions. Alpha controls fraction shuffled.""" g = gaps.copy() median_val = np.median(g) large_idx = np.where(g > median_val)[0] n_shuffle = max(2, int(alpha * len(large_idx))) chosen = rng.choice(large_idx, size=min(n_shuffle, len(large_idx)), replace=False) vals = g[chosen].copy() rng.shuffle(vals) g[chosen] = vals return g def perturb_uniform(gaps, alpha, rng): """Standard uniform partial shuffle. Baseline from previous runs.""" g = gaps.copy() n = len(g) n_swap = max(1, int(alpha * n / 2)) for _ in range(n_swap): i, j = rng.integers(0, n, size=2) g[i], g[j] = g[j], g[i] return g PERTURBATION_TYPES = { 'adjacent_swap': perturb_adjacent_swap, 'block_shuffle': perturb_block_shuffle, 'large_gap_only': perturb_large_gap_only, 'uniform': perturb_uniform, } # ---------- Main experiment ---------- def run_experiment(N=30000, seed=20260506, n_trials=16): rng = np.random.default_rng(seed) alphas = [0.1, 0.3, 0.5, 0.7, 0.9] obs_names = ['SR', 'L1', 'L2', 'SR2', 'triple_var'] results = {} for domain_name, gen_func in [('primes', lambda: generate_primes(N)), ('GUE', lambda: generate_gue(N, rng))]: gaps = gen_func() actual_n = len(gaps) print(f"\n=== {domain_name} (N={actual_n}) ===") original_obs = compute_observables(gaps) print(f"Original: {original_obs}") # Full shuffle baseline (z-score reference) full_shuffle_obs = {k: [] for k in obs_names} for _ in range(48): g_shuf = gaps.copy() rng.shuffle(g_shuf) obs = compute_observables(g_shuf) for k in obs_names: full_shuffle_obs[k].append(obs[k]) full_mean = {k: np.mean(v) for k, v in full_shuffle_obs.items()} full_std = {k: np.std(v) for k, v in full_shuffle_obs.items()} # For each perturbation type, collect observable retention curves # retention = (obs_perturbed - obs_shuffle) / (obs_original - obs_shuffle) domain_result = { 'original': original_obs, 'full_shuffle_mean': full_mean, 'full_shuffle_std': full_std, 'perturbations': {}, } # Collect: for each (perturbation_type, alpha) -> mean observable vector # Then do PCA across perturbation types all_profiles = [] # list of (pert_type, alpha, obs_vector) for pert_name, pert_func in PERTURBATION_TYPES.items(): pert_result = {} for alpha in alphas: trial_obs = {k: [] for k in obs_names} for t in range(n_trials): g_pert = pert_func(gaps, alpha, rng) obs = compute_observables(g_pert) for k in obs_names: trial_obs[k].append(obs[k]) mean_obs = {k: float(np.mean(v)) for k, v in trial_obs.items()} # Compute retention: how much of the original signal survives retention = {} for k in obs_names: denom = original_obs[k] - full_mean[k] if abs(denom) < 1e-15: retention[k] = 0.0 else: retention[k] = (mean_obs[k] - full_mean[k]) / denom pert_result[f'alpha_{alpha}'] = { 'mean': mean_obs, 'retention': retention, } obs_vec = [retention[k] for k in obs_names] all_profiles.append((pert_name, alpha, obs_vec)) domain_result['perturbations'][pert_name] = pert_result # PCA across all (perturbation_type, alpha) profiles exec /bin/bash -lc 'rg -n "def .*prime|primerange|sympy|generate.*prime|gue|hermitian" tools/exp_*.py tools/*.py' in /opt/MM_D-ND succeeded in 0ms: tools/exp_acf_stationarity.py:32:def sieve_primes(limit): tools/exp_acf_stationarity.py:42:def get_primes(n): tools/exp_acf_range_universality.py:21:from sympy import primerange tools/exp_acf_range_universality.py:24:def get_primes(n): tools/exp_acf_range_universality.py:29: primes = list(primerange(2, upper)) tools/exp_acf_range_universality.py:32: primes = list(primerange(2, upper)) tools/exp_acf_range_universality.py:170: gue_gaps = generate_rmt_spacings(N_rmt, n_mat_rmt, 'GUE') tools/exp_acf_range_universality.py:182: 'GUE': gue_gaps, tools/exp_3d_boundary_layers.py:27:def get_primes(n_max): tools/exp_3d_boundary_layers.py:36:def gue_gaps(N_mat, n_matrices, rng): tools/exp_3d_boundary_layers.py:183: gue_g = gue_gaps(n_mat, n_matrices, rng) tools/exp_3d_boundary_layers.py:184: if len(gue_g) > args.N: tools/exp_3d_boundary_layers.py:185: gue_g = gue_g[:args.N] tools/exp_3d_boundary_layers.py:186: gue_results, gue_orig, gue_bl_mean, gue_bl_std = run_crossover( tools/exp_3d_boundary_layers.py:187: gue_g, alphas, args.n_trials, rng, "GUE" tools/exp_3d_boundary_layers.py:210: ('gue', gue_results, gue_orig, gue_bl_mean, gue_bl_std), tools/exp_3d_boundary_layers.py:251: gue_sep = output['sequences']['gue']['layer_separation']['delta'] tools/exp_3d_boundary_layers.py:255: print(f"Layer separation Δα: Primes={prime_sep:+.3f}, GUE={gue_sep:+.3f}, Poisson={pois_sep:+.3f}") tools/exp_3d_boundary_layers.py:259: 'gue_layer_separation': float(gue_sep), tools/exp_beta_crossover.py:119:def gen_primes_multiscale(scales=None): tools/exp_acf_z6z_mechanism.py:24:def sieve_primes(n_max): tools/exp_acf_z6z_mechanism.py:34:def get_primes(n_primes): tools/exp_acf_amplitude_scaling.py:14:from sympy import primerange tools/exp_acf_amplitude_scaling.py:57: primes = np.array(list(primerange(2, upper)))[:args.n_primes] tools/exp_boundary_coherence.py:24:from sympy import primerange tools/exp_boundary_coherence.py:33: "spacing_ratio": {"poisson": 0.38629, "gue": 0.53590}, # 2ln2-1, 4-2√3 tools/exp_boundary_coherence.py:34: "gap_var_ratio": {"poisson": 1.0, "gue": 0.178}, tools/exp_boundary_coherence.py:35: "small_gap_frac": {"poisson": 0.2592, "gue": 0.020}, # P(s<0.3) for exp vs Wigner tools/exp_boundary_coherence.py:36: "brody_beta": {"poisson": 0.0, "gue": 1.0}, tools/exp_boundary_coherence.py:37: "lag1_acf": {"poisson": 0.0, "gue": -0.271}, tools/exp_boundary_coherence.py:115: g = REF[obs_name]["gue"] tools/exp_boundary_coherence.py:121:def generate_gue_spacings(n, n_matrices=50): tools/exp_boundary_coherence.py:141:def get_prime_gaps(pmin, pmax): tools/exp_boundary_coherence.py:143: primes = np.array(list(primerange(pmin, pmax))) tools/exp_boundary_coherence.py:162: gue_gaps = generate_gue_spacings(20000) tools/exp_boundary_coherence.py:163: gue_obs = compute_all_observables(gue_gaps) tools/exp_boundary_coherence.py:165: "raw": gue_obs, tools/exp_boundary_coherence.py:166: "tau": {k: to_tau(k, v) for k, v in gue_obs.items()}, tools/exp_boundary_growth.py:21:from sympy import primerange tools/exp_boundary_growth.py:60: primes = np.array(list(primerange(2, LIMIT)), dtype=np.int64) tools/exp_alpha_stability.py:15:from sympy import primerange tools/exp_alpha_stability.py:18:def get_primes(n): tools/exp_alpha_stability.py:25: primes = list(primerange(2, upper)) tools/exp_alpha_stability.py:28: primes = list(primerange(2, upper)) tools/exp_brody_crossover.py:25:def sieve_primes(limit): tools/exp_boundary_shuffle_audit.py:70:def gen_primes(n=100000): tools/exp_boundary_shuffle_audit.py:82:def gen_gue_eigenvalues(size=2000, n_matrices=50): tools/exp_boundary_shuffle_audit.py:244: 'gue': ('GUE random matrix', gen_gue_eigenvalues), tools/exp_boundary_shuffle_audit.py:275: dist_gue = abs(res['r_original'] - R_GUE) tools/exp_boundary_shuffle_audit.py:277: res['class_original'] = 'GUE' if dist_gue < dist_poi else 'Poisson' tools/exp_boundary_shuffle_audit.py:279: dist_gue_s = abs(res['r_shuffled_mean'] - R_GUE) tools/exp_boundary_shuffle_audit.py:281: res['class_shuffled'] = 'GUE' if dist_gue_s < dist_poi_s else 'Poisson' tools/exp_brody_calibration.py:109:def generate_primes(n_max=200000): tools/exp_brody_calibration.py:118:def prime_gaps_unfolded(n_gaps): tools/exp_brody_calibration.py:119: primes = generate_primes(n_gaps * 20)[:n_gaps + 1] tools/exp_brody_calibration.py:126:def gue_gaps(n_gaps, rng): tools/exp_brody_calibration.py:212: gue_g = gue_gaps(min(args.n_gaps, 400), rng) tools/exp_brody_calibration.py:213: obs_g = compute_observables(gue_g, n_shuffles=args.n_shuffles, rng=rng) tools/exp_brody_calibration.py:215: real_domains['gue_matrix'] = {**obs_g, 'beta_eff': beta_eff_g} tools/exp_brody_calibration.py:216: print(f"{'gue_matrix':>20} {obs_g['r']:8.4f} {obs_g['r_shuf']:8.4f} " tools/exp_cross_domain_dipolar_direction.py:28:def get_primes(n_max): tools/exp_coherence_length.py:23:def sieve_primes(limit): tools/exp_coherence_length.py:104:def measure_coherence_by_scale(all_gaps, all_primes, window_lengths, tools/exp_boundary_gue_poisson.py:22:from sympy import primerange tools/exp_boundary_gue_poisson.py:33:def primes_in_window(start, end, primes_array): tools/exp_boundary_gue_poisson.py:39:def cramer_random_primes(N_max, rng): tools/exp_boundary_gue_poisson.py:50:def analyze_windows(primes_array, windows): tools/exp_boundary_gue_poisson.py:69: primes = np.array(list(primerange(2, N_MAX))) tools/exp_boundary_gue_poisson.py:100: r_gue = 0.5307 # GOE (real symmetric) in 1D tools/exp_boundary_gue_poisson.py:120: print(f"\nReference: _GUE = {r_gue:.4f}, _Poisson = {r_poisson:.4f}") tools/exp_boundary_gue_poisson.py:161: dist_gue = abs(rp - r_gue) tools/exp_boundary_gue_poisson.py:163: label = "GUE" if dist_gue < dist_poi else "POISSON" tools/exp_boundary_gue_poisson.py:164: margin = abs(dist_gue - dist_poi) tools/exp_boundary_gue_poisson.py:171: "experiment": "boundary_gue_poisson_cramer", tools/exp_boundary_gue_poisson.py:176: "reference": {"r_gue": r_gue, "r_poisson": r_poisson}, tools/exp_coherence_robustness.py:5:Segue agent_20260416_0330 (COHERENCE_LENGTH). Obiettivo: stimare confidence intervals tools/exp_coherence_robustness.py:27:def sieve_primes(limit): tools/exp_dipolar_crossover.py:32:def gue_spacings(N_mat, n_matrices, rng): tools/exp_dipolar_crossover.py:49:def get_primes(n_max): tools/exp_dipolar_crossover.py:99: gue_mats = gue_spacings(N_mat, n_matrices, rng) tools/exp_dipolar_crossover.py:102: sr0, l1_0, _, _ = compute_dipolar(gue_mats) tools/exp_dipolar_crossover.py:106: for s in gue_mats: tools/exp_dipolar_crossover.py:121: trial_mats = [partial_shuffle(s, alpha, rng_trial) for s in gue_mats] tools/exp_crossover_phase_test.py:87:def generate_gue_gaps(N, rng): tools/exp_crossover_phase_test.py:102:def generate_prime_gaps(N): tools/exp_crossover_phase_test.py:104: import sympy tools/exp_crossover_phase_test.py:106: primes = list(sympy.primerange(2, limit)) tools/exp_crossover_phase_test.py:108: primes = list(sympy.primerange(2, limit * 2)) tools/exp_crossover_phase_test.py:216: sequences['GUE'] = generate_gue_gaps(args.N, rng) tools/exp_crossover_phase_test.py:218: sequences['Primes'] = generate_prime_gaps(args.N) tools/exp_brody_flow.py:26:def sieve_primes(n_max): tools/exp_magnitude_psd_from_acf.py:34:def sieve_primes(limit): tools/exp_magnitude_psd_from_acf.py:43:def get_primes(n_target): tools/exp_magnitude_psd_from_acf.py:53:def decompose_magnitude(primes): tools/exp_det_drift.py:22:from sympy import nextprime tools/exp_det_drift.py:26:def generate_primes(n_primes, start=2): tools/exp_det_drift.py:66: primes = generate_primes(window_size + 1, start=s) tools/exp_markov_layer_recovery_audit.py:89:def build_controls(prime_gaps, rng): tools/exp_markov_layer_recovery_audit.py:101: "gaps": generate_markov_surrogate(prime_gaps, 1, rng=rng), tools/exp_markov_layer_recovery_audit.py:105: "gaps": generate_markov_surrogate(prime_gaps, 2, rng=rng), tools/exp_dipolar_angle_reference.py:24:def get_primes(n_max): tools/exp_dipolar_angle_reference.py:79:def generate_gue_gaps(n_gaps, matrix_size=500): tools/exp_dipolar_angle_reference.py:120:def generate_cramer_gaps(primes): tools/exp_dipolar_angle_reference.py:155: gue_thetas = [] tools/exp_dipolar_angle_reference.py:156: gue_data = [] tools/exp_dipolar_angle_reference.py:158: gue_gaps = generate_gue_gaps(len(prime_gaps)) tools/exp_dipolar_angle_reference.py:159: theta, dsr, dl1, sr, l1, srs, l1s = dipolar_angle(gue_gaps, n_shuffle=50) tools/exp_dipolar_angle_reference.py:160: gue_thetas.append(theta) tools/exp_dipolar_angle_reference.py:161: gue_data.append((theta, dsr, dl1, sr, l1)) tools/exp_dipolar_angle_reference.py:164: gue_thetas = np.array(gue_thetas) tools/exp_dipolar_angle_reference.py:166: 'theta_mean': np.mean(gue_thetas), 'theta_std': np.std(gue_thetas), tools/exp_dipolar_angle_reference.py:167: 'thetas': gue_thetas.tolist(), tools/exp_dipolar_angle_reference.py:168: 'SR_mean': np.mean([d[3] for d in gue_data]), tools/exp_dipolar_angle_reference.py:169: 'L1_mean': np.mean([d[4] for d in gue_data]), tools/exp_dipolar_angle_reference.py:171: print(f" GUE: theta = {np.mean(gue_thetas):.1f} +/- {np.std(gue_thetas):.1f} deg") tools/exp_dipolar_angle_reference.py:178: goe_gaps = generate_goe_gaps(len(prime_gaps)) tools/exp_dipolar_angle_reference.py:198: poi_gaps = generate_poisson_gaps(len(prime_gaps)) tools/exp_dipolar_angle_reference.py:218: cramer_gaps = generate_cramer_gaps(primes_filtered) tools/exp_cross_observable_consistency.py:67:from sympy import primerange tools/exp_cross_observable_consistency.py:70:primes = np.array(list(primerange(2, PRIME_LIMIT)), dtype=float) tools/exp_cross_observable_consistency.py:89:def unfold_primes(p): tools/exp_cross_observable_consistency.py:108:def gue_gaps(n_eigenvalues=2000, n_matrices=5): tools/exp_cross_observable_consistency.py:171:gue_g = gue_gaps(n_eigenvalues=1500, n_matrices=4) tools/exp_cross_observable_consistency.py:172:r_gue = r_statistic(gue_g) tools/exp_cross_observable_consistency.py:173:beta_r_gue = beta_from_r(r_gue) tools/exp_cross_observable_consistency.py:174:print(f"r = {r_gue:.6f} → β_r = {beta_r_gue:.3f}") tools/exp_cross_observable_consistency.py:190:beta_sig_gue = {} tools/exp_cross_observable_consistency.py:195: beta_sig_gue[L] = b tools/exp_cross_observable_consistency.py:216:vals_gue = [f"{beta_sig_gue[L]:.3f}" for L in L_values] tools/exp_cross_observable_consistency.py:217:print(f"{'GUE':<12} {beta_r_gue:>6.3f} | " + " | ".join(f"{v:>9}" for v in vals_gue)) tools/exp_cross_observable_consistency.py:222:disagree_gue = max(abs(beta_r_gue - beta_sig_gue[L]) for L in L_values) tools/exp_cross_observable_consistency.py:227:print(f" GUE: {disagree_gue:.3f}") tools/exp_cross_observable_consistency.py:254: "gue": { tools/exp_cross_observable_consistency.py:255: "r": float(r_gue), tools/exp_cross_observable_consistency.py:256: "beta_r": float(beta_r_gue), tools/exp_cross_observable_consistency.py:257: "beta_sigma": {str(L): float(beta_sig_gue[L]) for L in L_values}, tools/exp_cross_observable_consistency.py:258: "max_disagreement": float(disagree_gue), tools/exp_excess_scaling.py:9:- Generate primes up to 10^8 using sympy sieve tools/exp_excess_scaling.py:19:from sympy import sieve tools/exp_meta_tautology_test.py:22:from sympy import primerange tools/exp_meta_tautology_test.py:25:def get_primes(n_max): tools/exp_meta_tautology_test.py:26: return np.array(list(primerange(2, n_max + 1)), dtype=np.int64) tools/exp_meta_tautology_test.py:131:def run(n_primes_max=600000, n_trials=20): tools/exp_geodesic_deviation_primes.py:21:from sympy import primerange tools/exp_geodesic_deviation_primes.py:26:primes = np.array(list(primerange(2, 10_000_000)), dtype=np.float64) tools/exp_desitter_unification.py:197:def z_score(prime_val, null_vals): tools/exp_markov_dipolar_decomposition.py:31:def get_primes(n_max): tools/exp_mod3_vs_residual_ordering.py:31:def sieve_primes(n_max): tools/exp_mod3_vs_residual_ordering.py:149:def cramer_random_primes(n_primes, primes_ref): tools/exp_markov_k_direction.py:32:def get_primes(n_max): tools/exp_dipolar_vector_scaling.py:21:from sympy import primerange tools/exp_dipolar_vector_scaling.py:24:def get_primes_in_range(lo, hi): tools/exp_dipolar_vector_scaling.py:26: return np.array(list(primerange(lo, hi)), dtype=np.int64) tools/exp_dipolar_vector_scaling.py:72:def analyze_scale(primes, label=""): tools/exp_observable_rank_audit.py:24:from exp_3d_boundary_layers import get_primes, gue_gaps, partial_shuffle tools/exp_observable_rank_audit.py:163: gue = gue_gaps(160, max(8, n_gaps // 160 + 1), rng).astype(float) tools/exp_observable_rank_audit.py:164: gue = gue[:n_gaps] tools/exp_observable_rank_audit.py:169: "gue": gue, tools/exp_markov_psd_prediction.py:25:def get_primes_sieve(n_max): tools/exp_markov_psd_prediction.py:36:def residue_sequence(primes): tools/exp_psd_prime_gaps.py:18:from sympy import primerange tools/exp_psd_prime_gaps.py:46:def run_experiment(n_primes=500_000, n_shuffles=20, nperseg=4096): tools/exp_psd_prime_gaps.py:52: primes = np.array(list(primerange(2, upper)))[:n_primes] tools/exp_markov_scale_function.py:27:def sieve_primes(limit): tools/exp_markov3_observable_hunt.py:26:def get_primes(n_max): tools/exp_selective_layer_decoupling.py:41:def get_primes(n_max): tools/exp_selective_layer_decoupling.py:50:def gen_prime_gaps(N): tools/exp_selective_layer_decoupling.py:55:def gen_gue_spacings(N, rng): tools/exp_selective_layer_decoupling.py:265: 'GUE': gen_gue_spacings(args.N, rng), tools/exp_modular_memory_spectrum.py:32:def sieve_primes(n_max): tools/exp_modular_memory_spectrum.py:42:def cramer_random_primes(n_max, rng): tools/exp_mod3_scaling.py:24:from sympy import primerange tools/exp_mod3_scaling.py:28:def get_primes_and_gaps(n_max): tools/exp_mod3_scaling.py:29: primes = np.array(list(primerange(2, n_max + 1))) tools/exp_markov_memory_by_gue_type.py:102:def generate_large_primes(n_limit=200000): tools/exp_markov_memory_by_gue_type.py:118:def generate_gue_gaps(n=2000): tools/exp_markov_memory_by_gue_type.py:139: 'gaps': generate_large_primes(200000), tools/exp_markov_memory_by_gue_type.py:143: domains['gue_matrix'] = { tools/exp_markov_memory_by_gue_type.py:144: 'gaps': generate_gue_gaps(3000), tools/exp_markov_memory_by_gue_type.py:208: 'gue_type': info['type'], tools/exp_markov_memory_by_gue_type.py:242: for r in sorted(results, key=lambda x: x['gue_type']): tools/exp_markov_memory_by_gue_type.py:243: print(f"{r['domain']:<22} {r['gue_type']:<18} {r['N']:>6} " tools/exp_markov_memory_by_gue_type.py:253: subset = [r for r in results if r['gue_type'] == gtype] tools/exp_markov_memory_by_gue_type.py:265: 'experiment': 'markov_memory_by_gue_type', tools/exp_markov_memory_by_gue_type.py:270: outpath = '/opt/MM_D-ND/tools/data/markov_memory_by_gue_type.json' tools/exp_two_channel_boundary.py:37:def sieve_primes(limit): tools/exp_two_channel_boundary.py:47:def get_primes(n_target): tools/exp_two_channel_boundary.py:89:def analyze_window(primes_window): tools/exp_two_channel_boundary.py:120:def analyze_shuffled(primes_window, rng): tools/exp_two_channel_boundary.py:151:def run(n_primes=500000, window=5000, n_surrogates=20): tools/exp_number_variance.py:15:from sympy import primerange tools/exp_number_variance.py:30:def unfolded_primes(primes): tools/exp_number_variance.py:59:def number_variance_gue(L_values): tools/exp_number_variance.py:73: primes = list(primerange(p_start, p_end)) tools/exp_number_variance.py:94: sv_gue = number_variance_gue(L_VALUES) tools/exp_number_variance.py:101: coeffs_gue = np.polyfit(log_L, sv_gue, 1) tools/exp_number_variance.py:128: print(f" {L:4d} | {sv_primes[i]:8.4f} | {sv_shuffled[i]:8.4f} | {sv_gue[i]:7.4f} | {sv_poisson[i]:8.1f}") tools/exp_number_variance.py:170: "gue_log_slope": float(2/np.pi**2), tools/exp_poisson_convergence.py:20:from sympy import primerange tools/exp_poisson_convergence.py:25:def get_primes(n_max): tools/exp_poisson_convergence.py:27: return np.array(list(primerange(2, n_max)), dtype=np.float64) tools/exp_poisson_convergence.py:71:def cramer_random_primes(primes, n_surrogates=10): tools/exp_poisson_convergence.py:81:def measure_window(primes_window): tools/exp_poisson_convergence.py:103:def measure_cramer_window(primes_window, n_surrogates=10): tools/exp_poisson_convergence.py:131:def run_experiment(n_primes_target=6_000_000, n_windows=25, n_surrogates=10, window_size=50_000): tools/exp_two_channel_shuffle_audit.py:17:from sympy import primerange tools/exp_two_channel_shuffle_audit.py:21:def get_prime_gaps(N): tools/exp_two_channel_shuffle_audit.py:23: primes = list(primerange(2, int(N * np.log(N) * 1.3) + 1000))[:N + 1] tools/exp_psd_amplitude_scaling.py:20:from sympy import primerange tools/exp_psd_amplitude_scaling.py:69: primes = np.array(list(primerange(2, upper)))[:args.n_primes] tools/exp_modular_algebra_depth.py:25:from sympy import primerange tools/exp_modular_algebra_depth.py:29: primes = list(primerange(2, N * 20))[:N+1] tools/exp_ricci_primes.py:20:from sympy import primerange tools/exp_ricci_primes.py:26:primes = np.array(list(primerange(2, 10_000_000)), dtype=np.float64) tools/exp_scale_selective_perturbation.py:31:from sympy import nextprime tools/exp_scale_selective_perturbation.py:34:def generate_primes(N): tools/exp_scale_selective_perturbation.py:47:def generate_gue(N, rng): tools/exp_scale_selective_perturbation.py:196: for domain_name, gen_func in [('primes', lambda: generate_primes(N)), tools/exp_scale_selective_perturbation.py:197: ('GUE', lambda: generate_gue(N, rng))]: tools/confine_spessore.py:14:from sympy import nextprime tools/confine_spessore.py:245: dist_gue = abs(r_mean - 0.5307) tools/confine_spessore.py:248: regime = "GUE" if dist_gue < dist_poi else "Poisson" tools/confine_spessore.py:251: f"dist_GUE={dist_gue:.4f} dist_Poi={dist_poi:.4f} " tools/exp_spectral_landscape.py:28:def gen_primes(n_spacings): tools/exp_spectral_landscape.py:51:def gen_gue(n_spacings): tools/exp_spectral_landscape.py:127: s_chaotic = gen_gue(n_chaotic) if n_chaotic > 100 else np.array([]) tools/exp_spectral_landscape.py:164: return gen_gue(n_spacings) # theoretically identical tools/exp_spectral_landscape.py:328: ("GUE_matrix", gen_gue, {}), tools/exp_perturbation_dimensionality_audit.py:30:def prime_gaps(n_gaps: int) -> np.ndarray: tools/exp_perturbation_dimensionality_audit.py:44:def gue_spacings(matrix_size: int, n_matrices: int, rng: np.random.Generator) -> np.ndarray: tools/exp_perturbation_dimensionality_audit.py:320: "gue_replicates": [], tools/exp_perturbation_dimensionality_audit.py:321: "gue_summary": {}, tools/exp_perturbation_dimensionality_audit.py:322: "gue_short_replicates": [], tools/exp_perturbation_dimensionality_audit.py:323: "gue_short_summary": {}, tools/exp_perturbation_dimensionality_audit.py:337: for i in range(args.n_gue_replicates): tools/exp_perturbation_dimensionality_audit.py:339: gaps = gue_spacings(args.gue_matrix_size, args.gue_matrices, rng) tools/exp_perturbation_dimensionality_audit.py:340: res = analyze(f"gue_rep_{i}", gaps, alphas, args.n_trials, args.n_baseline, rng) tools/exp_perturbation_dimensionality_audit.py:341: output["gue_replicates"].append(res) tools/exp_perturbation_dimensionality_audit.py:344: print(f"gue_rep_{i:<14} {len(gaps):>7} {res['pca']['effective_rank']:>7.3f} {pc2:>7.3f} {cos:>15.3f}") tools/exp_perturbation_dimensionality_audit.py:346: for i in range(args.n_gue_replicates): tools/exp_perturbation_dimensionality_audit.py:348: gaps = gue_spacings(args.gue_short_matrix_size, 1, rng) tools/exp_perturbation_dimensionality_audit.py:349: res = analyze(f"gue_short_rep_{i}", gaps, alphas, args.n_trials, args.n_baseline, rng) tools/exp_perturbation_dimensionality_audit.py:350: output["gue_short_replicates"].append(res) tools/exp_perturbation_dimensionality_audit.py:352: output["gue_summary"] = replicate_summary(output["gue_replicates"]) tools/exp_perturbation_dimensionality_audit.py:353: output["gue_short_summary"] = replicate_summary(output["gue_short_replicates"]) tools/exp_perturbation_dimensionality_audit.py:356: for label, summary in [("gue", output["gue_summary"]), ("gue_short", output["gue_short_summary"])]: tools/exp_perturbation_dimensionality_audit.py:375: parser.add_argument("--gue-matrix-size", type=int, default=180) tools/exp_perturbation_dimensionality_audit.py:376: parser.add_argument("--gue-matrices", type=int, default=16) tools/exp_perturbation_dimensionality_audit.py:377: parser.add_argument("--gue-short-matrix-size", type=int, default=42) tools/exp_perturbation_dimensionality_audit.py:378: parser.add_argument("--n-gue-replicates", type=int, default=6) tools/exp_spectral_rigidity.py:46:def generate_gue_gaps(n=600): tools/exp_spectral_rigidity.py:86: ('gue_matrix', {'gen': lambda: generate_gue_gaps(600), 'type': 'dist-GUE'}), tools/exp_spectral_rigidity.py:96: gue_theory = (2.0 / np.pi**2) * np.log(L_values) + 0.44 tools/exp_spectral_rigidity.py:164: 'sig2_gue_theory': gue_theory.tolist(), tools/exp_ricci_desitter_0406.py:4:from sympy import primerange tools/exp_ricci_desitter_0406.py:6:primes = np.array(list(primerange(2, 500_000)), dtype=np.float64) tools/exp_two_channel_decomposition.py:31:def sieve_primes(limit): tools/exp_two_channel_decomposition.py:40:def get_primes(n_target): tools/exp_two_channel_decomposition.py:51:def decompose_channels(primes): tools/exp_two_channel_decomposition.py:188:def scaling_analysis(primes, n_windows=15, n_surrogates=10): tools/exp_two_channel_psd.py:28:def sieve_primes(limit): tools/exp_two_channel_psd.py:37:def get_primes(n_target): tools/exp_two_channel_psd.py:47:def decompose_to_additive(primes): tools/exp_spectral_2d.py:70:def gen_primes(n_spacings): tools/exp_spectral_2d.py:88:def gen_primes_raw_gaps(n_spacings): tools/exp_spectral_2d.py:102:def gen_gue(n_spacings): tools/exp_spectral_2d.py:115: s_chaotic = gen_gue(n_chaotic) if n_chaotic > 50 else np.random.exponential(1.0, n_chaotic) tools/exp_spectral_2d.py:327: ("GUE", lambda n: gen_gue(n)), tools/exp_two_layer_universality.py:35:def get_primes(n_max): tools/exp_two_layer_universality.py:183:def gen_prime_gaps(N): tools/exp_two_layer_universality.py:188:def gen_gue_spacings(N, rng=None): tools/exp_two_layer_universality.py:249: 'GUE': gen_gue_spacings, tools/build_lab_graph.py:129: 'title_en': title, # i report hanno titoli tecnici — stesso in entrambe le lingue tools/build_lab_graph.py:702: if 'gue' in full or 'poisson' in full: teorie_report.update(['Q', 'G']) tools/exp_two_channel_cross_domain.py:31: python tools/exp_two_channel_cross_domain.py [--n_primes N] [--gue_size N] [--n_windows N] tools/exp_two_channel_cross_domain.py:41:def sieve_primes(limit): tools/exp_two_channel_cross_domain.py:51:def get_primes(n_target): tools/exp_two_channel_cross_domain.py:62:def gue_eigenvalues(n_matrices, matrix_size): tools/exp_two_channel_cross_domain.py:85:def cramer_random_primes(n_primes): tools/exp_two_channel_cross_domain.py:117:def decompose_primes(primes_window): tools/exp_two_channel_cross_domain.py:209:def analyze_at_scale_primes(primes_window, n_surrogates=30, rng=None): tools/exp_two_channel_cross_domain.py:266:def run(n_primes=200000, gue_matrices=50, gue_size=800, n_windows=8, window=5000, n_surrogates=20): tools/exp_two_channel_cross_domain.py:296: print(f"\n=== GUE EIGENVALUES ({gue_matrices} matrices of size {gue_size}) ===") tools/exp_two_channel_cross_domain.py:297: gue_spacings = gue_eigenvalues(gue_matrices, gue_size) tools/exp_two_channel_cross_domain.py:298: print(f"Got {len(gue_spacings)} GUE spacings, mean={np.mean(gue_spacings):.3f}") tools/exp_two_channel_cross_domain.py:301: gue_max_start = len(gue_spacings) - window - 10 tools/exp_two_channel_cross_domain.py:302: gue_starts = np.unique(np.logspace(1, np.log10(max(gue_max_start, 100)), n_windows).astype(int)) tools/exp_two_channel_cross_domain.py:303: gue_starts = gue_starts[gue_starts < gue_max_start] tools/exp_two_channel_cross_domain.py:305: gue_results = [] tools/exp_two_channel_cross_domain.py:306: for s in gue_starts: tools/exp_two_channel_cross_domain.py:307: gw = gue_spacings[s:s + window] tools/exp_two_channel_cross_domain.py:314: gue_results.append(obs) tools/exp_two_channel_cross_domain.py:319: results['gue'] = gue_results tools/exp_two_channel_cross_domain.py:418: parser.add_argument('--gue_matrices', type=int, default=50) tools/exp_two_channel_cross_domain.py:419: parser.add_argument('--gue_size', type=int, default=800) tools/exp_two_channel_cross_domain.py:425: results = run(args.n_primes, args.gue_matrices, args.gue_size, tools/exp_two_channel_cross_domain.py:437: 'gue_matrices': args.gue_matrices, tools/exp_two_channel_cross_domain.py:438: 'gue_size': args.gue_size, tools/dnd_experiments.py:4:Esegue gli esperimenti generati dall'engine. tools/dnd_experiments.py:515: gue = [(d, sp, r) for d, sp, r, _ in domains if sp > GUE_THRESHOLD] tools/dnd_experiments.py:518: print(f"\n GUE ({len(gue)} domini):") tools/dnd_experiments.py:519: for d, sp, r in gue: tools/dnd_experiments.py:527: r_gue = [r for _, _, r in gue] tools/dnd_experiments.py:530: if r_gue and r_poisson: tools/dnd_experiments.py:531: mean_gue = np.mean(r_gue) tools/dnd_experiments.py:533: std_gue = np.std(r_gue) if len(r_gue) > 1 else 0 tools/dnd_experiments.py:537: gue_range = (min(r_gue), max(r_gue)) tools/dnd_experiments.py:539: overlap = max(0, min(gue_range[1], poisson_range[1]) - max(gue_range[0], poisson_range[0])) tools/dnd_experiments.py:540: total_range = max(gue_range[1], poisson_range[1]) - min(gue_range[0], poisson_range[0]) tools/dnd_experiments.py:546: print(f" r_diretto GUE: {mean_gue:.4f} ± {std_gue:.4f} range=[{gue_range[0]:.4f}, {gue_range[1]:.4f}]") tools/dnd_experiments.py:558: 'gue': {'n': len(gue), 'r_mean': mean_gue, 'r_std': std_gue, 'range': gue_range}, tools/dnd_experiments.py:1153:# RUNNER — Esegue tutti gli esperimenti e salva risultati tools/dnd_experiments.py:1157: """Esegue tutti gli esperimenti e aggiorna lo stato.""" tools/exp_two_channel_universality.py:42:def sieve_primes(limit): tools/exp_two_channel_universality.py:51:def get_primes(n_target): tools/dnd_implications.py:17: distingue automaticamente i due casi. tools/dnd_implications.py:39: Il file .git/hooks/pre-push esegue --guard su ogni file modificato tools/dnd_arxiv.py:182: Esegue uno scan arXiv. tools/dipartimento.py:6:2. Esegue esperimenti computazionali (autoricerca + tools) tools/dipartimento.py:119: 'test': 'zeta_spacing_gue', tools/dipartimento.py:215: Per ogni claim, esegue il test e riporta PASS/FAIL/SKIP. tools/dipartimento.py:255: """Esegue un singolo test di verifica.""" tools/dipartimento.py:311: elif test_name == 'zeta_spacing_gue': tools/dipartimento.py:330: from sympy import primerange tools/dipartimento.py:331: primes = list(primerange(2, 10000)) tools/dipartimento.py:962: gue = [e for e in journal if e.get('spacing') == 'GUE-like' and isinstance(e.get('ciclo'), int)] tools/dipartimento.py:964: if gue and poisson: tools/dipartimento.py:968: 'claim': f'{len(gue)} domini GUE, {len(poisson)} Poisson — il confine è il terzo incluso operativo', tools/dipartimento.py:1356: Se i risultati mancano, li genera. Se falliscono, logga e prosegue. tools/dipartimento.py:1403: """Esegue zeta_validation senza figure (solo dati).""" tools/dipartimento.py:1412: """Esegue topological_charge senza figure (solo dati).""" tools/dipartimento.py:1429: """Esegue bloch_explorer con scan ridotto (solo dati, no figure).""" tools/dipartimento.py:1658: for kw in ('primi', 'prime', 'gap', 'acf', 'brody', 'poisson', 'gue', 'spacing')) tools/dnd_lab.py:590: """Esegue tutti i banchi di prova.""" tools/dnd_lab.py:1245: """Il numero di gap nel Cantor set segue la sequenza di Fibonacci?""" tools/dnd_lab.py:1393: 3. Se RISONANTE o PARZIALE: esegue l'esperimento tools/dnd_lab.py:1426: 'GUE': 'random_matrix_gue', tools/dnd_curva.py:9: Cosa distingue il punto (1,-1) dagli altri punti sulla curva? tools/dnd_curva.py:198:# PARTE 2: Cosa distingue k=1? — Le tre misure tools/dnd_curva.py:204: Ma cosa distingue k=1? tools/dnd_M_operator.py:61: 'question_template': '{claim} e\' conseguenza di det=-1 o specifico di phi?', tools/dnd_M_operator.py:174: elif any(w in claim for w in ('spacing', '', 'gue', 'poisson')): tools/dnd_autoricerca.py:653: 'gue_dist': abs(mean_r - 0.5996), tools/dnd_autoricerca.py:829: if spacing.get('tipo') == 'GUE-like' and spacing.get('gue_dist', 1) < 0.1: tools/dnd_autoricerca.py:831: 'tipo': 'spacing_gue', tools/dnd_autoricerca.py:1152: gue_domains = [e['dominio'] for e in journal if e.get('spacing') == 'GUE-like'] tools/dnd_autoricerca.py:1154: print(f" GUE-like: {gue_domains}") tools/dnd_autoricerca.py:1158: if len(gue_domains) > 0 and len(poisson_domains) > 0: tools/dnd_autoricerca.py:1159: ratio_cluster = len(gue_domains) / len(poisson_domains) tools/dnd_autoricerca.py:1331: gue = [e for e in reali if e.get('spacing') == 'GUE-like' and e.get('spacing_r')] tools/dnd_autoricerca.py:1334: for g in gue: tools/dnd_autoricerca.py:1345: gue_rs = [e['r_diretto'] for e in gue if e.get('r_diretto')] tools/dnd_autoricerca.py:1347: if gue_rs and poi_rs: tools/dnd_autoricerca.py:1348: mean_gue = np.mean(gue_rs) tools/dnd_autoricerca.py:1350: ratio = mean_gue / mean_poi if mean_poi > 0 else float('inf') tools/dnd_autoricerca.py:1351: print(f" GUE mean r_diretto: {mean_gue:.4f}") tools/dnd_autoricerca.py:1516: A13 — La seconda voce. Prosegue dal punto in cui il seme e' arrivato. tools/dnd_autoricerca.py:1718: 'gue_domains': [], # domini con spacing GUE tools/dnd_autoricerca.py:1726: campo['gue_domains'].append(entry.get('dominio', '?')) tools/dnd_autoricerca.py:1734: if f.get('tipo') in ('spacing_gue', 'struttura_dnd_piena', 'convergenza_triviale'): tools/dnd_autoricerca.py:1887: n_gue = len(campo['gue_domains']) tools/dnd_autoricerca.py:1889: if n_gue + n_poi > 0: tools/dnd_autoricerca.py:1890: print(f"\n Campo vivo dopo Fase 0: {n_gue} GUE / {n_poi} Poisson") tools/dnd_autoricerca.py:1891: report_lines.append(f"\n Campo dopo Fase 0: {n_gue} GUE / {n_poi} Poisson") tools/dnd_autoricerca.py:2012: n_gue = len(campo['gue_domains']) tools/dnd_autoricerca.py:2017: report_lines.append(f" GUE: {n_gue} | Poisson: {n_poi} | Vincoli: {n_vinc} | Anomalie: {n_anom}") tools/dnd_autoricerca.py:2019: report_lines.append(f" Domini GUE: {', '.join(campo['gue_domains'][:10])}") tools/dnd_autoricerca.py:2024: print(f"\n Campo vivo finale: {n_gue} GUE / {n_poi} Poisson / {n_vinc} vincoli / {n_anom} anomalie / ⟨r⟩={avg_r:.4f}") tools/dnd_autoricerca.py:2064: 'gue': n_gue, 'poisson': n_poi, tools/dnd_autoricerca.py:2069: 'domini_gue': campo['gue_domains'][:10], tools/dnd_autoricerca.py:2122: f in e.get('findings', []) for f in ['rapporto_aureo_diretto', 'spacing_gue'] tools/dnd_autoricerca.py:2387: from sympy import primerange tools/dnd_autoricerca.py:2388: primes = np.array(list(primerange(2, max_n)), dtype=float) tools/dnd_autoricerca.py:2402: from sympy import primerange tools/dnd_autoricerca.py:2403: primes = np.array(list(primerange(2, max_n)), dtype=float) tools/dnd_banchi_tm1.py:10:Non esegue calcoli pesanti — prende risultati dal domandatore e li tools/dnd_domandatore.py:378: """Esegue un singolo esperimento e cattura output.""" tools/dnd_domandatore.py:990: Non esegue esperimenti — genera le 5 domande strutturali che tools/dnd_quantization.py:41:def V_prime(r): tools/dnd_quantization.py:130: def V_eff_prime(r): tools/dnd_quantization.py:166: def V_map_prime(x): tools/dnd_quantization.py:315: d_gue = abs(stats['mean_r'] - 0.5996) tools/dnd_quantization.py:317: closer = "GUE" if d_gue < d_poi else "Poisson" tools/dnd_quantization.py:329: d_gue = abs(stats_h['mean_r'] - 0.5996) tools/dnd_quantization.py:333: f"→ {'GUE' if d_gue < d_poi else 'Poi'}") tools/dnd_compatibility.py:149: "random_matrix_gue", tools/dnd_compatibility.py:150: "GUE (Gaussian Unitary Ensemble): matrici hermitiane casuali, " tools/dnd_dipolo_lab.py:11:- Segue le assonanze tools/dnd_dipolo_lab.py:125: """Esegue il domandatore su una tensione.""" tools/dnd_dipolo_lab.py:163: Esploratore. Lancia traiettorie, genera domande, segue assonanze. tools/dnd_indeterminazione.py:270: def V_prime(r): tools/dnd_indeterminazione.py:551: │ La compressione segue un potenziale V(r) con tre │ tools/dnd_scenario.py:141: La risultante R(t) segue la dinamica F(R) = 1/R + 1 - R. tools/dnd_scenario.py:921: è il prossimo passo. La traiettoria segue le assonanze convergenti. tools/dnd_scenario.py:983: Segue la traiettoria lagrangiana: il percorso naturale tools/dnd_trace_bridge.py:145: d_gue = abs(stats['mean_r'] - 0.5996) tools/dnd_trace_bridge.py:147: closer = "GUE" if d_gue < d_poi else "Poisson" tools/dnd_trace_bridge.py:354: d_gue = abs(val['mean_r'] - 0.5996) tools/dnd_trace_bridge.py:356: marker = "***" if d_gue < 0.03 else "" tools/dnd_trace_bridge.py:357: print(f" {key:40s} ={val['mean_r']:.4f} {'GUE' if d_gue < d_poi else 'Poi'} {marker}") tools/dnd_next.py:378: Esegue la decisione. Ciclo chiuso: decide → agisci → verifica → aggiorna seme. tools/dnd_next.py:425: """Esegue CRYSTALLIZE: scoperta confermata → lab experiment → seed update.""" tools/dnd_next.py:477: Esegue EXPLORE: il sistema ragiona sulla tensione e decide cosa fare. tools/dnd_next.py:481: Il sistema lo esegue, valuta, e produce la prossima tensione. tools/dnd_next.py:535: # claude -p è un agente: ragiona E esegue. Non generiamo codice separato. tools/dnd_next.py:538: "Sei un ricercatore autonomo. Hai accesso a numpy, scipy, sympy. " tools/dnd_next.py:596: """Esegue RESEARCH: direzione dal seme.""" tools/dnd_next.py:606: """Esegue INTEGRATE: risultati freschi → segnala quali paper aggiornare.""" tools/dnd_engine.py:115: 'test': 'test_zeta_gue', tools/dnd_engine.py:332: 'motivo': 'Verifica diretta per n=1..20. det(M^n)=(-1)^n. Conseguenza di M²=M+I (Cayley-Hamilton).', tools/dnd_engine.py:340: 'motivo': 'Verifica diretta per n=1..30. F_31/F_30 = phi a 13 cifre. Conseguenza di cf(phi)=[1;1,1,...].', tools/dnd_engine.py:364: 'motivo': 'Conseguenza della struttura di M mod p in GL(2,F_p). Il periodo divide |GL(2,F_p)| = p(p-1)(p+1)(p-1).', tools/dnd_engine.py:420: 'motivo': 'Conseguenza di cf=[1;1,...]. A N non-Fibonacci: 3 gap con ENTRAMBI i rapporti = phi. Nessun altro irrazionale ha questa proprieta.', tools/dnd_engine.py:478: 'motivo': 'test_spacings() di dnd_gue_test.py applicato agli spacing dei rapporti F_{n+1}/F_n. Classifica automaticamente la statistica.', tools/dnd_engine.py:479: 'test': 'test_ext_spacing_gue', tools/dnd_engine.py:567:def test_zeta_gue(): tools/dnd_engine.py:580: 'is_gue': 0.5 < sr < 0.7, tools/dnd_engine.py:666: gue_standard = 0.60 tools/dnd_engine.py:668: 'pass': sr > gue_standard + 0.1, tools/dnd_engine.py:670: 'excess_over_gue': float(sr - gue_standard), tools/dnd_engine.py:671: 'anomalous': sr > gue_standard + 0.1, tools/dnd_engine.py:1267: from sympy import isprime tools/dnd_engine.py:1384: Silver = log(1+sqrt(2)). Il Lyapunov on-surface distingue topologicamente.""" tools/dnd_engine.py:1908:def test_ext_spacing_gue(): tools/dnd_engine.py:1929: def gue_cdf(s): tools/dnd_engine.py:1936: ks_gue, p_gue = kstest(spacings, gue_cdf) tools/dnd_engine.py:1940: fits = [("GUE", ks_gue, p_gue), ("GOE", ks_goe, p_goe), ("Poisson", ks_poi, p_poi)] tools/dnd_engine.py:1948: 'ks_gue': float(ks_gue), tools/dnd_zero_controllo.py:10: a scale grandi, possiamo distinguerli. Se restano sovrapposti, non tools/dnd_zero_controllo.py:38:def gap_random_cramer(n_primes, seed=42): tools/dnd_lab_team.py:169:3. Se l'assunzione è falsa, cosa ne consegue? tools/dnd_paper_audit.py:426: suggestion=f'Remove "{word}" and provide explicit argument, or qualify with "we argue that..."' tools/dnd_paper_audit.py:429: # 2. Vague quantifiers tools/dnd_paper_audit.py:430: vague = re.findall(r'\b(?:many|several|some|various|numerous)\s+(?:authors|works|studies|results)', tools/dnd_paper_audit.py:432: if len(vague) > 3: tools/dnd_paper_audit.py:435: f'{len(vague)} vague quantifiers ("many authors", "several works", etc.)', tools/dnd_paper_audit.py:846: """Esegue l'audit completo su un paper.""" tools/dnd_zero_operator.py:105: """Esegue l'esperimento Zero su un dominio.""" tools/dnd_gue_test.py:39:def gue_cdf(s): tools/dnd_gue_test.py:261: for t_guess in candidates: tools/dnd_gue_test.py:265: t_zero = float(mpmath.findroot(f, t_guess).real) tools/dnd_gue_test.py:272: idx = np.argmin(np.abs(t_values - t_guess)) tools/dnd_gue_test.py:274: zeros.append(float(t_guess)) tools/dnd_gue_test.py:366: ks_gue, p_gue = kstest(spacings, gue_cdf) tools/dnd_gue_test.py:374: fits = [("GUE", ks_gue, p_gue), ("GOE", ks_goe, p_goe), ("Poisson", ks_poi, p_poi)] tools/dnd_gue_test.py:382: "ks_gue": {"KS": float(ks_gue), "p": float(p_gue)}, tools/dnd_gue_test.py:546: gue_count = 0 tools/dnd_gue_test.py:557: gue_count += 1 tools/dnd_gue_test.py:567: elif gue_count == total: tools/dnd_gue_test.py:568: verdict = (f"GUE UNIVERSALE: {gue_count}/{total} L-functions mostrano GUE. " tools/dnd_gue_test.py:571: elif gue_count >= total // 2: tools/dnd_gue_test.py:572: verdict = (f"GUE DOMINANTE: {gue_count}/{total}. Risultato parziale, " tools/dnd_gue_test.py:631: beta_gue = sum(1 for lb in beta_results.values() for v in lb.values() if v.get("class") == "GUE") tools/dnd_gue_test.py:633: print(f" DIPOLO FRATTALE: {beta_gue}/{beta_total} GUE a livello di repulsione") tools/dnd_gue_test.py:687: all_neg1_gue = all(b > 1.5 for b in betas_neg1) tools/dnd_gue_test.py:688: all_pos1_gue = all(b > 1.5 for b in betas_pos1) tools/dnd_gue_test.py:690: if all_neg1_gue and not all_pos1_gue: tools/dnd_gue_test.py:694: elif all_neg1_gue and all_pos1_gue: tools/dnd_gue_test.py:698: elif not all_neg1_gue: tools/dnd_gue_test.py:703: family_verdict = f"MISTO: neg1 GUE={all_neg1_gue}, pos1 GUE={all_pos1_gue}" tools/dnd_gue_test.py:727: if beta_gue == beta_total and beta_total > 0: tools/dnd_gue_test.py:728: verdict_fractal = (f"GUE CONFERMATO via dipolo frattale: β>2 per {beta_gue}/{beta_total} misure. " tools/dnd_gue_test.py:729: f"KS globale={gue_count}/{total} GUE (forma bulk). " tools/dnd_gue_test.py:730: f"β allo zero={beta_gue}/{beta_total} GUE (repulsione). " tools/dnd_gue_test.py:734: verdict_fractal = (f"Dipolo frattale: {beta_gue}/{beta_total} GUE. " tools/dnd_gue_test.py:735: f"KS globale: {gue_count}/{total}. Risultato misto.") tools/dnd_gue_test.py:789: "gue_count_ks": gue_count, tools/dnd_gue_test.py:790: "gue_count_beta": beta_gue, tools/dnd_gue_test.py:812: outpath = os.path.join(os.path.dirname(__file__), "data", "piano11b_gue_test.json") tools/dnd_zeros_vs_zeta.py:122:def gue_spacings(n=1000, matrix_size=200): tools/dnd_zeros_vs_zeta.py:217: s_gue = gue_spacings(2000) tools/dnd_zeros_vs_zeta.py:220: print(f" GUE: {len(s_gue)} spacings") tools/dnd_zeros_vs_zeta.py:233: ks_zg, p_zg = ks_2samp(s_zeta, s_gue) tools/dnd_zeros_vs_zeta.py:243: ks_dg, p_dg = ks_2samp(res["spacings"], s_gue) tools/dnd_zeros_vs_zeta.py:247: r_gue = spacing_ratio_statistic(s_gue) tools/dnd_zeros_vs_zeta.py:248: ks_gg, p_gg = ks_2samp(s_gue, s_zeta) tools/dnd_zeros_vs_zeta.py:249: print(f"{'GUE (RMT)':<25} {r_gue:<8.4f} {ks_gg:<12.4f} {p_gg:<12.4e} {'---':<12} {'---':<12}") tools/dnd_zeros_vs_zeta.py:253: ks_pg, p_pg = ks_2samp(s_poisson, s_gue) tools/dnd_zeros_vs_zeta.py:258: ks_sg, p_sg = ks_2samp(s_synth, s_gue) tools/dnd_zeros_vs_zeta.py:325: ax3.hist(s_gue, bins=bins, density=True, alpha=0.3, label='GUE (RMT)', color='green') tools/dnd_zeros_vs_zeta.py:348: r_values = [r_zeta, r_gue, r_poisson, r_synth] tools/dnd_zeros_vs_zeta.py:408: gue_dist = abs(r_best - 0.5996) tools/dnd_zeros_vs_zeta.py:413: print(f" Distance to GUE: {gue_dist:.4f}") tools/dnd_zeros_vs_zeta.py:417: if gue_dist < poisson_dist: tools/dnd_zeros_vs_zeta.py:424: return dnd_results, s_zeta, s_gue tools/dnd_projective_quantization.py:206: # dove V_i segue la sequenza di Fibonacci: V_i = V·χ(⌊(i+1)/φ⌋ - ⌊i/φ⌋) tools/dnd_projective_quantization.py:255: d_gue = abs(mean_r - 0.5996) tools/dnd_projective_quantization.py:257: closer = "GUE" if d_gue < d_poi else "Poisson" tools/dnd_projective_quantization.py:263: if d_crit < min(d_gue, d_poi): tools/dnd_projective_quantization.py:318: H_gue = np.random.randn(N, N) + 1j * np.random.randn(N, N) tools/dnd_projective_quantization.py:319: H_gue = (H_gue + H_gue.conj().T) / 2 tools/dnd_projective_quantization.py:320: eigs_gue = np.sort(np.real(linalg.eigvalsh(H_gue))) tools/dnd_projective_quantization.py:321: spacing_stats(eigs_gue, f"GUE N={N}") tools/dnd_loop.py:391: """Esegue un ciclo completo su un topic.""" tools/exp_acf_stationarity.py:32:def sieve_primes(limit): tools/exp_acf_stationarity.py:42:def get_primes(n): tools/dnd_piano11.py:149:def gue_cdf(s): tools/dnd_piano11.py:161:def gue_pdf(s): tools/dnd_piano11.py:205: ks_gue = ks_test_against(norm_spacings, gue_cdf, "GUE") tools/dnd_piano11.py:209: best = min([ks_gue, ks_goe, ks_poi], key=lambda x: x["KS"]) tools/dnd_piano11.py:215: "ks_gue": ks_gue, tools/dnd_piano11.py:249: ks_zeta_gue = ks_test_against(zeta_norm, gue_cdf, "GUE") tools/dnd_piano11.py:250: ks_L_gue = ks_test_against(L_norm, gue_cdf, "GUE") tools/dnd_piano11.py:262: hist_gue = gue_pdf(bin_centers) tools/dnd_piano11.py:265: l2_L_gue = float(np.sqrt(np.sum((hist_L - hist_gue)**2) * ds)) tools/dnd_piano11.py:266: l2_zeta_gue = float(np.sqrt(np.sum((hist_zeta - hist_gue)**2) * ds)) tools/dnd_piano11.py:272: "ks_L_vs_gue": ks_L_gue, tools/dnd_piano11.py:273: "ks_zeta_vs_gue": ks_zeta_gue, tools/dnd_piano11.py:279: "L_to_gue": l2_L_gue, tools/dnd_piano11.py:280: "zeta_to_gue": l2_zeta_gue, tools/dnd_piano11.py:287: "gue_theory": hist_gue.tolist() tools/dnd_piano11.py:289: "same_universality_class": bool(ks_L_gue["p"] > 0.05 and ks_zeta_gue["p"] > 0.05) tools/dnd_piano11.py:541: print(f" KS vs GUE: {p2['ks_gue']['KS']:.4f} (p={p2['ks_gue']['p']:.4f})") tools/dnd_piano11.py:554: print(f" L(s,χ₅) vs GUE: KS={p3['ks_L_vs_gue']['KS']:.4f}") tools/dnd_piano11.py:555: print(f" ζ vs GUE: KS={p3['ks_zeta_vs_gue']['KS']:.4f}") tools/exp_boundary_coherence.py:24:from sympy import primerange tools/exp_boundary_coherence.py:33: "spacing_ratio": {"poisson": 0.38629, "gue": 0.53590}, # 2ln2-1, 4-2√3 tools/exp_boundary_coherence.py:34: "gap_var_ratio": {"poisson": 1.0, "gue": 0.178}, tools/exp_boundary_coherence.py:35: "small_gap_frac": {"poisson": 0.2592, "gue": 0.020}, # P(s<0.3) for exp vs Wigner tools/exp_boundary_coherence.py:36: "brody_beta": {"poisson": 0.0, "gue": 1.0}, tools/exp_boundary_coherence.py:37: "lag1_acf": {"poisson": 0.0, "gue": -0.271}, tools/exp_boundary_coherence.py:115: g = REF[obs_name]["gue"] tools/exp_boundary_coherence.py:121:def generate_gue_spacings(n, n_matrices=50): tools/exp_boundary_coherence.py:141:def get_prime_gaps(pmin, pmax): tools/exp_boundary_coherence.py:143: primes = np.array(list(primerange(pmin, pmax))) tools/exp_boundary_coherence.py:162: gue_gaps = generate_gue_spacings(20000) tools/exp_boundary_coherence.py:163: gue_obs = compute_all_observables(gue_gaps) tools/exp_boundary_coherence.py:165: "raw": gue_obs, tools/exp_boundary_coherence.py:166: "tau": {k: to_tau(k, v) for k, v in gue_obs.items()}, tools/dnd_research_engine.py:11:5. EXECUTOR: esegue, raccoglie risultati tools/dnd_research_engine.py:492: 'gue_dist': abs(mean_r - 0.5996), tools/dnd_research_engine.py:546: 'type': 'gue_like_spacing', tools/dnd_research_engine.py:548: 'gue_dist': result['gue_dist'], tools/exp_brody_calibration.py:109:def generate_primes(n_max=200000): tools/exp_brody_calibration.py:118:def prime_gaps_unfolded(n_gaps): tools/exp_brody_calibration.py:119: primes = generate_primes(n_gaps * 20)[:n_gaps + 1] tools/exp_brody_calibration.py:126:def gue_gaps(n_gaps, rng): tools/exp_brody_calibration.py:212: gue_g = gue_gaps(min(args.n_gaps, 400), rng) tools/exp_brody_calibration.py:213: obs_g = compute_observables(gue_g, n_shuffles=args.n_shuffles, rng=rng) tools/exp_brody_calibration.py:215: real_domains['gue_matrix'] = {**obs_g, 'beta_eff': beta_eff_g} tools/exp_brody_calibration.py:216: print(f"{'gue_matrix':>20} {obs_g['r']:8.4f} {obs_g['r_shuf']:8.4f} " tools/dnd_teoria.py:137: 'campo': 'Einstein: non puoi distinguerli localmente'}, tools/exp_coherence_robustness.py:5:Segue agent_20260416_0330 (COHERENCE_LENGTH). Obiettivo: stimare confidence intervals tools/exp_coherence_robustness.py:27:def sieve_primes(limit): tools/dnd_riformulazioni.py:663: """Esegue tutte le riformulazioni.""" tools/exp_dipolar_angle_reference.py:24:def get_primes(n_max): tools/exp_dipolar_angle_reference.py:79:def generate_gue_gaps(n_gaps, matrix_size=500): tools/exp_dipolar_angle_reference.py:120:def generate_cramer_gaps(primes): tools/exp_dipolar_angle_reference.py:155: gue_thetas = [] tools/exp_dipolar_angle_reference.py:156: gue_data = [] tools/exp_dipolar_angle_reference.py:158: gue_gaps = generate_gue_gaps(len(prime_gaps)) tools/exp_dipolar_angle_reference.py:159: theta, dsr, dl1, sr, l1, srs, l1s = dipolar_angle(gue_gaps, n_shuffle=50) tools/exp_dipolar_angle_reference.py:160: gue_thetas.append(theta) tools/exp_dipolar_angle_reference.py:161: gue_data.append((theta, dsr, dl1, sr, l1)) tools/exp_dipolar_angle_reference.py:164: gue_thetas = np.array(gue_thetas) tools/exp_dipolar_angle_reference.py:166: 'theta_mean': np.mean(gue_thetas), 'theta_std': np.std(gue_thetas), tools/exp_dipolar_angle_reference.py:167: 'thetas': gue_thetas.tolist(), tools/exp_dipolar_angle_reference.py:168: 'SR_mean': np.mean([d[3] for d in gue_data]), tools/exp_dipolar_angle_reference.py:169: 'L1_mean': np.mean([d[4] for d in gue_data]), tools/exp_dipolar_angle_reference.py:171: print(f" GUE: theta = {np.mean(gue_thetas):.1f} +/- {np.std(gue_thetas):.1f} deg") tools/exp_dipolar_angle_reference.py:178: goe_gaps = generate_goe_gaps(len(prime_gaps)) tools/exp_dipolar_angle_reference.py:198: poi_gaps = generate_poisson_gaps(len(prime_gaps)) tools/exp_dipolar_angle_reference.py:218: cramer_gaps = generate_cramer_gaps(primes_filtered) tools/dnd_spettro_zeta.py:220:def generate_gue_spacings(n_samples=50000): tools/dnd_spettro_zeta.py:271: s_gue_ref = generate_gue_spacings(50000) tools/dnd_spettro_zeta.py:277: ks_gue, p_gue = stats.ks_2samp(s_data, s_gue_ref) tools/dnd_spettro_zeta.py:282: print(f" vs GUE: KS={ks_gue:.4f}, p={p_gue:.4f}") tools/dnd_spettro_zeta.py:288: 'vs_GUE': {'KS': float(ks_gue), 'p': float(p_gue)}, tools/dnd_spettro_zeta.py:728: var_gue = 0.178 tools/dnd_spettro_zeta.py:762: dist_gue = abs(var_dnd - var_gue) tools/dnd_spettro_zeta.py:763: if dist_goe < dist_gue: tools/dnd_spectral_probe.py:51:def gue_cdf(s): tools/dnd_spectral_probe.py:252: ks_gue, p_gue = kstest(self.spacings, gue_cdf) tools/dnd_spectral_probe.py:256: fits = [("GUE", ks_gue, p_gue), ("GOE", ks_goe, p_goe), ("Poisson", ks_poi, p_poi)] tools/dnd_spectral_probe.py:263: "ks_gue": {"KS": round(float(ks_gue), 4), "p": round(float(p_gue), 4)}, tools/dnd_spectral_probe.py:1057: all_gue = all(b > 1.5 for b in valid) tools/dnd_spectral_probe.py:1058: print(f" All GUE: {all_gue} ({sum(1 for b in valid if b > 1.5)}/{len(valid)})") tools/exp_geodesic_deviation_primes.py:21:from sympy import primerange tools/exp_geodesic_deviation_primes.py:26:primes = np.array(list(primerange(2, 10_000_000)), dtype=np.float64) tools/exp_acf_amplitude_scaling.py:14:from sympy import primerange tools/exp_acf_amplitude_scaling.py:57: primes = np.array(list(primerange(2, upper)))[:args.n_primes] tools/dnd_torre.py:116: # La conseguenza: OGNI invariante è f(tr, det) tools/dnd_torre.py:117: print(f"\n Conseguenza: per 2×2, gli invarianti indipendenti sono ESATTAMENTE 2:") tools/dnd_torre.py:174: # Dopo, tutto segue dalla ricorrenza L(n) = L(n-1) + L(n-2) tools/exp_markov_k_direction.py:32:def get_primes(n_max): tools/exp_alpha_stability.py:15:from sympy import primerange tools/exp_alpha_stability.py:18:def get_primes(n): tools/exp_alpha_stability.py:25: primes = list(primerange(2, upper)) tools/exp_alpha_stability.py:28: primes = list(primerange(2, upper)) tools/dnd_trace_bridge_v3.py:20:Domanda: la distribuzione delle frequenze di Δ distingue ζ da random? tools/dnd_trace_bridge_v3.py:179: d_gue = abs(stats['mean_r'] - 0.5996) tools/dnd_trace_bridge_v3.py:181: closer = "GUE" if d_gue < d_poi else "Poisson" tools/dnd_trace_bridge_v3.py:203: d_gue = abs(r_val - 0.5996) tools/dnd_trace_bridge_v3.py:205: closer = "GUE" if d_gue < d_poi else "Poi" tools/dnd_trace_bridge_v3.py:333: d_gue = abs(stats_rz['mean_r'] - 0.5996) tools/dnd_trace_bridge_v3.py:335: print(f" Spacing dei ratio (zeta): ={stats_rz['mean_r']:.4f} → {'GUE' if d_gue < d_poi else 'Poi'}") tools/dnd_trace_bridge_v3.py:337: d_gue = abs(stats_rr['mean_r'] - 0.5996) tools/dnd_trace_bridge_v3.py:339: print(f" Spacing dei ratio (random): ={stats_rr['mean_r']:.4f} → {'GUE' if d_gue < d_poi else 'Poi'}") tools/dnd_zero_notturno.py:300: """Esegue tutte le 5 misure su scale crescenti.""" tools/exp_markov_scale_function.py:27:def sieve_primes(limit): tools/exp_boundary_gue_poisson.py:22:from sympy import primerange tools/exp_boundary_gue_poisson.py:33:def primes_in_window(start, end, primes_array): tools/exp_boundary_gue_poisson.py:39:def cramer_random_primes(N_max, rng): tools/exp_boundary_gue_poisson.py:50:def analyze_windows(primes_array, windows): tools/exp_boundary_gue_poisson.py:69: primes = np.array(list(primerange(2, N_MAX))) tools/exp_boundary_gue_poisson.py:100: r_gue = 0.5307 # GOE (real symmetric) in 1D tools/exp_boundary_gue_poisson.py:120: print(f"\nReference: _GUE = {r_gue:.4f}, _Poisson = {r_poisson:.4f}") tools/exp_boundary_gue_poisson.py:161: dist_gue = abs(rp - r_gue) tools/exp_boundary_gue_poisson.py:163: label = "GUE" if dist_gue < dist_poi else "POISSON" tools/exp_boundary_gue_poisson.py:164: margin = abs(dist_gue - dist_poi) tools/exp_boundary_gue_poisson.py:171: "experiment": "boundary_gue_poisson_cramer", tools/exp_boundary_gue_poisson.py:176: "reference": {"r_gue": r_gue, "r_poisson": r_poisson}, tools/dnd_zero_controllo2.py:86:def gap_cramer(n_primes, seed=42): tools/exp_mod3_scaling.py:24:from sympy import primerange tools/exp_mod3_scaling.py:28:def get_primes_and_gaps(n_max): tools/exp_mod3_scaling.py:29: primes = np.array(list(primerange(2, n_max + 1))) tools/exp_brody_flow.py:26:def sieve_primes(n_max): tools/exp_acf_range_universality.py:21:from sympy import primerange tools/exp_acf_range_universality.py:24:def get_primes(n): tools/exp_acf_range_universality.py:29: primes = list(primerange(2, upper)) tools/exp_acf_range_universality.py:32: primes = list(primerange(2, upper)) tools/exp_acf_range_universality.py:170: gue_gaps = generate_rmt_spacings(N_rmt, n_mat_rmt, 'GUE') tools/exp_acf_range_universality.py:182: 'GUE': gue_gaps, tools/dnd_zero_traiettoria.py:154: print(f" STRUTTURA: la distanza da phi segue una legge di potenza") tools/exp_number_variance.py:15:from sympy import primerange tools/exp_number_variance.py:30:def unfolded_primes(primes): tools/exp_number_variance.py:59:def number_variance_gue(L_values): tools/exp_number_variance.py:73: primes = list(primerange(p_start, p_end)) tools/exp_number_variance.py:94: sv_gue = number_variance_gue(L_VALUES) tools/exp_number_variance.py:101: coeffs_gue = np.polyfit(log_L, sv_gue, 1) tools/exp_number_variance.py:128: print(f" {L:4d} | {sv_primes[i]:8.4f} | {sv_shuffled[i]:8.4f} | {sv_gue[i]:7.4f} | {sv_poisson[i]:8.1f}") tools/exp_number_variance.py:170: "gue_log_slope": float(2/np.pi**2), tools/exp_beta_crossover.py:119:def gen_primes_multiscale(scales=None): tools/exp_3d_boundary_layers.py:27:def get_primes(n_max): tools/exp_3d_boundary_layers.py:36:def gue_gaps(N_mat, n_matrices, rng): tools/exp_3d_boundary_layers.py:183: gue_g = gue_gaps(n_mat, n_matrices, rng) tools/exp_3d_boundary_layers.py:184: if len(gue_g) > args.N: tools/exp_3d_boundary_layers.py:185: gue_g = gue_g[:args.N] tools/exp_3d_boundary_layers.py:186: gue_results, gue_orig, gue_bl_mean, gue_bl_std = run_crossover( tools/exp_3d_boundary_layers.py:187: gue_g, alphas, args.n_trials, rng, "GUE" tools/exp_3d_boundary_layers.py:210: ('gue', gue_results, gue_orig, gue_bl_mean, gue_bl_std), tools/exp_3d_boundary_layers.py:251: gue_sep = output['sequences']['gue']['layer_separation']['delta'] tools/exp_3d_boundary_layers.py:255: print(f"Layer separation Δα: Primes={prime_sep:+.3f}, GUE={gue_sep:+.3f}, Poisson={pois_sep:+.3f}") tools/exp_3d_boundary_layers.py:259: 'gue_layer_separation': float(gue_sep), tools/exp_psd_amplitude_scaling.py:20:from sympy import primerange tools/exp_psd_amplitude_scaling.py:69: primes = np.array(list(primerange(2, upper)))[:args.n_primes] tools/exp_boundary_shuffle_audit.py:70:def gen_primes(n=100000): tools/exp_boundary_shuffle_audit.py:82:def gen_gue_eigenvalues(size=2000, n_matrices=50): tools/exp_boundary_shuffle_audit.py:244: 'gue': ('GUE random matrix', gen_gue_eigenvalues), tools/exp_boundary_shuffle_audit.py:275: dist_gue = abs(res['r_original'] - R_GUE) tools/exp_boundary_shuffle_audit.py:277: res['class_original'] = 'GUE' if dist_gue < dist_poi else 'Poisson' tools/exp_boundary_shuffle_audit.py:279: dist_gue_s = abs(res['r_shuffled_mean'] - R_GUE) tools/exp_boundary_shuffle_audit.py:281: res['class_shuffled'] = 'GUE' if dist_gue_s < dist_poi_s else 'Poisson' tools/exp_cross_observable_consistency.py:67:from sympy import primerange tools/exp_cross_observable_consistency.py:70:primes = np.array(list(primerange(2, PRIME_LIMIT)), dtype=float) tools/exp_cross_observable_consistency.py:89:def unfold_primes(p): tools/exp_cross_observable_consistency.py:108:def gue_gaps(n_eigenvalues=2000, n_matrices=5): tools/exp_cross_observable_consistency.py:171:gue_g = gue_gaps(n_eigenvalues=1500, n_matrices=4) tools/exp_cross_observable_consistency.py:172:r_gue = r_statistic(gue_g) tools/exp_cross_observable_consistency.py:173:beta_r_gue = beta_from_r(r_gue) tools/exp_cross_observable_consistency.py:174:print(f"r = {r_gue:.6f} → β_r = {beta_r_gue:.3f}") tools/exp_cross_observable_consistency.py:190:beta_sig_gue = {} tools/exp_cross_observable_consistency.py:195: beta_sig_gue[L] = b tools/exp_cross_observable_consistency.py:216:vals_gue = [f"{beta_sig_gue[L]:.3f}" for L in L_values] tools/exp_cross_observable_consistency.py:217:print(f"{'GUE':<12} {beta_r_gue:>6.3f} | " + " | ".join(f"{v:>9}" for v in vals_gue)) tools/exp_cross_observable_consistency.py:222:disagree_gue = max(abs(beta_r_gue - beta_sig_gue[L]) for L in L_values) tools/exp_cross_observable_consistency.py:227:print(f" GUE: {disagree_gue:.3f}") tools/exp_cross_observable_consistency.py:254: "gue": { tools/exp_cross_observable_consistency.py:255: "r": float(r_gue), tools/exp_cross_observable_consistency.py:256: "beta_r": float(beta_r_gue), tools/exp_cross_observable_consistency.py:257: "beta_sigma": {str(L): float(beta_sig_gue[L]) for L in L_values}, tools/exp_cross_observable_consistency.py:258: "max_disagreement": float(disagree_gue), tools/exp_acf_z6z_mechanism.py:24:def sieve_primes(n_max): tools/exp_acf_z6z_mechanism.py:34:def get_primes(n_primes): tools/exp_coherence_length.py:23:def sieve_primes(limit): tools/exp_coherence_length.py:104:def measure_coherence_by_scale(all_gaps, all_primes, window_lengths, tools/exp_scale_selective_perturbation.py:31:from sympy import nextprime tools/exp_scale_selective_perturbation.py:34:def generate_primes(N): tools/exp_scale_selective_perturbation.py:47:def generate_gue(N, rng): tools/exp_scale_selective_perturbation.py:196: for domain_name, gen_func in [('primes', lambda: generate_primes(N)), tools/exp_scale_selective_perturbation.py:197: ('GUE', lambda: generate_gue(N, rng))]: tools/exp_desitter_unification.py:197:def z_score(prime_val, null_vals): tools/exp_spectral_rigidity.py:46:def generate_gue_gaps(n=600): tools/exp_spectral_rigidity.py:86: ('gue_matrix', {'gen': lambda: generate_gue_gaps(600), 'type': 'dist-GUE'}), tools/exp_spectral_rigidity.py:96: gue_theory = (2.0 / np.pi**2) * np.log(L_values) + 0.44 tools/exp_spectral_rigidity.py:164: 'sig2_gue_theory': gue_theory.tolist(), tools/exp_boundary_growth.py:21:from sympy import primerange tools/exp_boundary_growth.py:60: primes = np.array(list(primerange(2, LIMIT)), dtype=np.int64) tools/exp_dipolar_vector_scaling.py:21:from sympy import primerange tools/exp_dipolar_vector_scaling.py:24:def get_primes_in_range(lo, hi): tools/exp_dipolar_vector_scaling.py:26: return np.array(list(primerange(lo, hi)), dtype=np.int64) tools/exp_dipolar_vector_scaling.py:72:def analyze_scale(primes, label=""): tools/exp_two_channel_psd.py:28:def sieve_primes(limit): tools/exp_two_channel_psd.py:37:def get_primes(n_target): tools/exp_two_channel_psd.py:47:def decompose_to_additive(primes): tools/gap_ratio_primes.py:2:from sympy import primerange tools/gap_ratio_primes.py:3:primes = list(primerange(2, 7920))[:1000] tools/exp_markov3_observable_hunt.py:26:def get_primes(n_max): tools/exp_brody_crossover.py:25:def sieve_primes(limit): tools/exp_crossover_phase_test.py:87:def generate_gue_gaps(N, rng): tools/exp_crossover_phase_test.py:102:def generate_prime_gaps(N): tools/exp_crossover_phase_test.py:104: import sympy tools/exp_crossover_phase_test.py:106: primes = list(sympy.primerange(2, limit)) tools/exp_crossover_phase_test.py:108: primes = list(sympy.primerange(2, limit * 2)) tools/exp_crossover_phase_test.py:216: sequences['GUE'] = generate_gue_gaps(args.N, rng) tools/exp_crossover_phase_test.py:218: sequences['Primes'] = generate_prime_gaps(args.N) tools/lab_trajectory_apply.py:37: python3 lab_trajectory_apply.py --apply # esegue le modifiche tools/lab_trajectory_apply.py:336: help="Esegue le modifiche. Senza questo flag → dry-run (default).", tools/exp_cross_domain_dipolar_direction.py:28:def get_primes(n_max): tools/exp_det_drift.py:22:from sympy import nextprime tools/exp_det_drift.py:26:def generate_primes(n_primes, start=2): tools/exp_det_drift.py:66: primes = generate_primes(window_size + 1, start=s) tools/exp_excess_scaling.py:9:- Generate primes up to 10^8 using sympy sieve tools/exp_excess_scaling.py:19:from sympy import sieve tools/exp_dipolar_crossover.py:32:def gue_spacings(N_mat, n_matrices, rng): tools/exp_dipolar_crossover.py:49:def get_primes(n_max): tools/exp_dipolar_crossover.py:99: gue_mats = gue_spacings(N_mat, n_matrices, rng) tools/exp_dipolar_crossover.py:102: sr0, l1_0, _, _ = compute_dipolar(gue_mats) tools/exp_dipolar_crossover.py:106: for s in gue_mats: tools/exp_dipolar_crossover.py:121: trial_mats = [partial_shuffle(s, alpha, rng_trial) for s in gue_mats] tools/observables_registry.py:260: gue_like = rng.gamma(shape=2.0, scale=0.5, size=200) tools/observables_registry.py:261: res = compute_canonical(gue_like) tools/observables_registry.py:267: print(f" SR_local_rigidity = {SR_local_rigidity(gue_like):.6f}") tools/observables_registry.py:268: print(f" triple_var_normalized = {triple_var_normalized(gue_like):.6f}") tools/exp_markov_dipolar_decomposition.py:31:def get_primes(n_max): tools/exp_magnitude_psd_from_acf.py:34:def sieve_primes(limit): tools/exp_magnitude_psd_from_acf.py:43:def get_primes(n_target): tools/exp_magnitude_psd_from_acf.py:53:def decompose_magnitude(primes): tools/exp_markov_psd_prediction.py:25:def get_primes_sieve(n_max): tools/exp_markov_psd_prediction.py:36:def residue_sequence(primes): tools/exp_markov_memory_by_gue_type.py:102:def generate_large_primes(n_limit=200000): tools/exp_markov_memory_by_gue_type.py:118:def generate_gue_gaps(n=2000): tools/exp_markov_memory_by_gue_type.py:139: 'gaps': generate_large_primes(200000), tools/exp_markov_memory_by_gue_type.py:143: domains['gue_matrix'] = { tools/exp_markov_memory_by_gue_type.py:144: 'gaps': generate_gue_gaps(3000), tools/exp_markov_memory_by_gue_type.py:208: 'gue_type': info['type'], tools/exp_markov_memory_by_gue_type.py:242: for r in sorted(results, key=lambda x: x['gue_type']): tools/exp_markov_memory_by_gue_type.py:243: print(f"{r['domain']:<22} {r['gue_type']:<18} {r['N']:>6} " tools/exp_markov_memory_by_gue_type.py:253: subset = [r for r in results if r['gue_type'] == gtype] tools/exp_markov_memory_by_gue_type.py:265: 'experiment': 'markov_memory_by_gue_type', tools/exp_markov_memory_by_gue_type.py:270: outpath = '/opt/MM_D-ND/tools/data/markov_memory_by_gue_type.json' tools/exp_markov_layer_recovery_audit.py:89:def build_controls(prime_gaps, rng): tools/exp_markov_layer_recovery_audit.py:101: "gaps": generate_markov_surrogate(prime_gaps, 1, rng=rng), tools/exp_markov_layer_recovery_audit.py:105: "gaps": generate_markov_surrogate(prime_gaps, 2, rng=rng), tools/exp_meta_tautology_test.py:22:from sympy import primerange tools/exp_meta_tautology_test.py:25:def get_primes(n_max): tools/exp_meta_tautology_test.py:26: return np.array(list(primerange(2, n_max + 1)), dtype=np.int64) tools/exp_meta_tautology_test.py:131:def run(n_primes_max=600000, n_trials=20): tools/exp_modular_memory_spectrum.py:32:def sieve_primes(n_max): tools/exp_modular_memory_spectrum.py:42:def cramer_random_primes(n_max, rng): tools/exp_modular_algebra_depth.py:25:from sympy import primerange tools/exp_modular_algebra_depth.py:29: primes = list(primerange(2, N * 20))[:N+1] tools/exp_mod3_vs_residual_ordering.py:31:def sieve_primes(n_max): tools/exp_mod3_vs_residual_ordering.py:149:def cramer_random_primes(n_primes, primes_ref): tools/exp_poisson_convergence.py:20:from sympy import primerange tools/exp_poisson_convergence.py:25:def get_primes(n_max): tools/exp_poisson_convergence.py:27: return np.array(list(primerange(2, n_max)), dtype=np.float64) tools/exp_poisson_convergence.py:71:def cramer_random_primes(primes, n_surrogates=10): tools/exp_poisson_convergence.py:81:def measure_window(primes_window): tools/exp_poisson_convergence.py:103:def measure_cramer_window(primes_window, n_surrogates=10): tools/exp_poisson_convergence.py:131:def run_experiment(n_primes_target=6_000_000, n_windows=25, n_surrogates=10, window_size=50_000): tools/exp_observable_rank_audit.py:24:from exp_3d_boundary_layers import get_primes, gue_gaps, partial_shuffle tools/exp_observable_rank_audit.py:163: gue = gue_gaps(160, max(8, n_gaps // 160 + 1), rng).astype(float) tools/exp_observable_rank_audit.py:164: gue = gue[:n_gaps] tools/exp_observable_rank_audit.py:169: "gue": gue, tools/r_stat_primes.py:2:from sympy import primerange tools/r_stat_primes.py:5:primes = list(primerange(2, 3572))[:500] tools/r_stat_primes.py:16:r_gue = 4 - 2 * np.sqrt(3) # ~0.536 tools/r_stat_primes.py:25: "r_gue_theory": round(r_gue, 4), tools/r_stat_primes.py:27: "verdict": "GUE-like" if r_real > (r_poisson + r_gue) / 2 else "Poisson-like" tools/exp_perturbation_dimensionality_audit.py:30:def prime_gaps(n_gaps: int) -> np.ndarray: tools/exp_perturbation_dimensionality_audit.py:44:def gue_spacings(matrix_size: int, n_matrices: int, rng: np.random.Generator) -> np.ndarray: tools/exp_perturbation_dimensionality_audit.py:320: "gue_replicates": [], tools/exp_perturbation_dimensionality_audit.py:321: "gue_summary": {}, tools/exp_perturbation_dimensionality_audit.py:322: "gue_short_replicates": [], tools/exp_perturbation_dimensionality_audit.py:323: "gue_short_summary": {}, tools/exp_perturbation_dimensionality_audit.py:337: for i in range(args.n_gue_replicates): tools/exp_perturbation_dimensionality_audit.py:339: gaps = gue_spacings(args.gue_matrix_size, args.gue_matrices, rng) tools/exp_perturbation_dimensionality_audit.py:340: res = analyze(f"gue_rep_{i}", gaps, alphas, args.n_trials, args.n_baseline, rng) tools/exp_perturbation_dimensionality_audit.py:341: output["gue_replicates"].append(res) tools/exp_perturbation_dimensionality_audit.py:344: print(f"gue_rep_{i:<14} {len(gaps):>7} {res['pca']['effective_rank']:>7.3f} {pc2:>7.3f} {cos:>15.3f}") tools/exp_perturbation_dimensionality_audit.py:346: for i in range(args.n_gue_replicates): tools/exp_perturbation_dimensionality_audit.py:348: gaps = gue_spacings(args.gue_short_matrix_size, 1, rng) tools/exp_perturbation_dimensionality_audit.py:349: res = analyze(f"gue_short_rep_{i}", gaps, alphas, args.n_trials, args.n_baseline, rng) tools/exp_perturbation_dimensionality_audit.py:350: output["gue_short_replicates"].append(res) tools/exp_perturbation_dimensionality_audit.py:352: output["gue_summary"] = replicate_summary(output["gue_replicates"]) tools/exp_perturbation_dimensionality_audit.py:353: output["gue_short_summary"] = replicate_summary(output["gue_short_replicates"]) tools/exp_perturbation_dimensionality_audit.py:356: for label, summary in [("gue", output["gue_summary"]), ("gue_short", output["gue_short_summary"])]: tools/exp_perturbation_dimensionality_audit.py:375: parser.add_argument("--gue-matrix-size", type=int, default=180) tools/exp_perturbation_dimensionality_audit.py:376: parser.add_argument("--gue-matrices", type=int, default=16) tools/exp_perturbation_dimensionality_audit.py:377: parser.add_argument("--gue-short-matrix-size", type=int, default=42) tools/exp_perturbation_dimensionality_audit.py:378: parser.add_argument("--n-gue-replicates", type=int, default=6) tools/exp_ricci_primes.py:20:from sympy import primerange tools/exp_ricci_primes.py:26:primes = np.array(list(primerange(2, 10_000_000)), dtype=np.float64) tools/exp_psd_prime_gaps.py:18:from sympy import primerange tools/exp_psd_prime_gaps.py:46:def run_experiment(n_primes=500_000, n_shuffles=20, nperseg=4096): tools/exp_psd_prime_gaps.py:52: primes = np.array(list(primerange(2, upper)))[:n_primes] tools/exp_ricci_desitter_0406.py:4:from sympy import primerange tools/exp_ricci_desitter_0406.py:6:primes = np.array(list(primerange(2, 500_000)), dtype=np.float64) tools/spectral_gap_analysis.py:77:def _sieve_primes(n): tools/exp_spectral_landscape.py:28:def gen_primes(n_spacings): tools/exp_spectral_landscape.py:51:def gen_gue(n_spacings): tools/exp_spectral_landscape.py:127: s_chaotic = gen_gue(n_chaotic) if n_chaotic > 100 else np.array([]) tools/exp_spectral_landscape.py:164: return gen_gue(n_spacings) # theoretically identical tools/exp_spectral_landscape.py:328: ("GUE_matrix", gen_gue, {}), tools/exp_selective_layer_decoupling.py:41:def get_primes(n_max): tools/exp_selective_layer_decoupling.py:50:def gen_prime_gaps(N): tools/exp_selective_layer_decoupling.py:55:def gen_gue_spacings(N, rng): tools/exp_selective_layer_decoupling.py:265: 'GUE': gen_gue_spacings(args.N, rng), tools/exp_spectral_2d.py:70:def gen_primes(n_spacings): tools/exp_spectral_2d.py:88:def gen_primes_raw_gaps(n_spacings): tools/exp_spectral_2d.py:102:def gen_gue(n_spacings): tools/exp_spectral_2d.py:115: s_chaotic = gen_gue(n_chaotic) if n_chaotic > 50 else np.random.exponential(1.0, n_chaotic) tools/exp_spectral_2d.py:327: ("GUE", lambda n: gen_gue(n)), tools/exp_two_channel_decomposition.py:31:def sieve_primes(limit): tools/exp_two_channel_decomposition.py:40:def get_primes(n_target): tools/exp_two_channel_decomposition.py:51:def decompose_channels(primes): tools/exp_two_channel_decomposition.py:188:def scaling_analysis(primes, n_windows=15, n_surrogates=10): tools/exp_two_channel_boundary.py:37:def sieve_primes(limit): tools/exp_two_channel_boundary.py:47:def get_primes(n_target): tools/exp_two_channel_boundary.py:89:def analyze_window(primes_window): tools/exp_two_channel_boundary.py:120:def analyze_shuffled(primes_window, rng): tools/exp_two_channel_boundary.py:151:def run(n_primes=500000, window=5000, n_surrogates=20): tools/exp_two_layer_universality.py:35:def get_primes(n_max): tools/exp_two_layer_universality.py:183:def gen_prime_gaps(N): tools/exp_two_layer_universality.py:188:def gen_gue_spacings(N, rng=None): tools/exp_two_layer_universality.py:249: 'GUE': gen_gue_spacings, tools/exp_two_channel_cross_domain.py:31: python tools/exp_two_channel_cross_domain.py [--n_primes N] [--gue_size N] [--n_windows N] tools/exp_two_channel_cross_domain.py:41:def sieve_primes(limit): tools/exp_two_channel_cross_domain.py:51:def get_primes(n_target): tools/exp_two_channel_cross_domain.py:62:def gue_eigenvalues(n_matrices, matrix_size): tools/exp_two_channel_cross_domain.py:85:def cramer_random_primes(n_primes): tools/exp_two_channel_cross_domain.py:117:def decompose_primes(primes_window): tools/exp_two_channel_cross_domain.py:209:def analyze_at_scale_primes(primes_window, n_surrogates=30, rng=None): tools/exp_two_channel_cross_domain.py:266:def run(n_primes=200000, gue_matrices=50, gue_size=800, n_windows=8, window=5000, n_surrogates=20): tools/exp_two_channel_cross_domain.py:296: print(f"\n=== GUE EIGENVALUES ({gue_matrices} matrices of size {gue_size}) ===") tools/exp_two_channel_cross_domain.py:297: gue_spacings = gue_eigenvalues(gue_matrices, gue_size) tools/exp_two_channel_cross_domain.py:298: print(f"Got {len(gue_spacings)} GUE spacings, mean={np.mean(gue_spacings):.3f}") tools/exp_two_channel_cross_domain.py:301: gue_max_start = len(gue_spacings) - window - 10 tools/exp_two_channel_cross_domain.py:302: gue_starts = np.unique(np.logspace(1, np.log10(max(gue_max_start, 100)), n_windows).astype(int)) tools/exp_two_channel_cross_domain.py:303: gue_starts = gue_starts[gue_starts < gue_max_start] tools/exp_two_channel_cross_domain.py:305: gue_results = [] tools/exp_two_channel_cross_domain.py:306: for s in gue_starts: tools/exp_two_channel_cross_domain.py:307: gw = gue_spacings[s:s + window] tools/exp_two_channel_cross_domain.py:314: gue_results.append(obs) tools/exp_two_channel_cross_domain.py:319: results['gue'] = gue_results tools/exp_two_channel_cross_domain.py:418: parser.add_argument('--gue_matrices', type=int, default=50) tools/exp_two_channel_cross_domain.py:419: parser.add_argument('--gue_size', type=int, default=800) tools/exp_two_channel_cross_domain.py:425: results = run(args.n_primes, args.gue_matrices, args.gue_size, tools/exp_two_channel_cross_domain.py:437: 'gue_matrices': args.gue_matrices, tools/exp_two_channel_cross_domain.py:438: 'gue_size': args.gue_size, tools/exp_two_channel_shuffle_audit.py:17:from sympy import primerange tools/exp_two_channel_shuffle_audit.py:21:def get_prime_gaps(N): tools/exp_two_channel_shuffle_audit.py:23: primes = list(primerange(2, int(N * np.log(N) * 1.3) + 1000))[:N + 1] tools/exp_two_channel_universality.py:42:def sieve_primes(limit): tools/exp_two_channel_universality.py:51:def get_primes(n_target): tools/gue_gap_test.py:3:from sympy import primerange tools/gue_gap_test.py:5:primes = list(primerange(2, 3572))[:500] tools/gue_gap_test.py:34:gue_like = r_data > 0.5 tools/gue_gap_test.py:36:if gue_like and significativo: tools/gue_gap_test.py:38:elif gue_like and not significativo: tools/gue_gap_test.py:49: "gue_like": bool(gue_like), tools/lab_affinatore.py:104:Una riga sul dove il ciclo seguente potrebbe proseguire, se scegliesse. tools/lab_valutatore.py:89: produrle. Esegue il filtro, non genera il contenuto. tools/r_ratio_decay.py:8:Risultato: il decay segue una power law r - Poisson = A * N^(-alpha) con alpha ~ 0.06. tools/r_ratio_decay.py:18:from sympy import nextprime tools/r_ratio_decay.py:39:def generate_primes(N): tools/r_ratio_decay.py:53: primes = generate_primes(N_max) tools/riemann_R.py:12:import sympy as sp tools/riemann_R.py:13:from sympy import sqrt, Rational, I, oo, Matrix, symbols, series, simplify tools/riemann_R.py:14:from sympy import cos, sin, pi, exp, Function, limit, zoo tools/riemann_R.py:17:from sympy import nextprime tools/riemann_R.py:388: print("\n--- Verifica simbolica (sympy) per n=1..6 ---") tools/test_gue_poisson_boundary.py:18:gue_r = [] tools/test_gue_poisson_boundary.py:22: gue_r.append(spacing_ratio(eigvalsh(H))) tools/test_gue_poisson_boundary.py:28:mu_g, mu_p = np.mean(gue_r), np.mean(poi_r) tools/test_gue_poisson_boundary.py:32:clean_g = sum(1 for r in gue_r if r > boundary) / n_mat tools/zeta_validation.py:71:def create_emergence_system(N, eigenvalue_pattern='prime'): tools/zeta_validation.py:109:def _sieve_primes(n): tools/zeta_validation.py:227:def run_validation(n_zeros=100, N_system=100, eigenvalue_pattern='prime', exec /bin/bash -lc "sed -n '1,220p' tools/data/seme.json" in /opt/MM_D-ND succeeded in 0ms: { "timestamp": "2026-05-06T06:33:02.679931", "piano": 66, "tensioni": [ { "tipo": "confine_inesplorato", "id": "TRASCENDENZA_LIMITE", "claim": "La trascendenza e il limite attuale del modello. I punti fissi relazionali (non solo phi ma la rete di punti fissi tra osservabili) possono rivelare il vero grafo della realta e pattern nelle matrici. Il confine non e nella matematica - e nel passaggio tra piani.", "intensita": 0.9, "nota": "Input operatore 2026-04-10. Tocca: confine del modello, struttura relazionale dei punti fissi. Consecutio: quali punti fissi relazionali emergono dalle 21 tensioni attuali? Il grafo e gia nei dati?", "manuale": true, "porta": "sessione_interattiva", "condensato_ref": "A3,A10", "condensato_motivo": "Estende A3 (punto fisso singolo) a rete relazionale. Tocca A10 (dipolo) come caso speciale." }, { "tipo": "scoperta", "id": "DUALITA_DIPOLARE_VS_ILLUSORIA", "claim": "Due tipi di dualita: (1) dipolare - generativa, il modello (det=-1), (2) illusoria - dispersiva, entropia (det=+1). Le regole incoerenti producono la seconda. La dualita illusoria e entropia come dispersione, non come informazione.", "intensita": 0.9, "nota": "Input operatore 2026-04-10. Tocca: entropia come dispersione illusoria vs generazione dipolare. Consecutio: nel Lab i domini Poisson (entropia massima) mostrano dualita illusoria? I domini GUE (strutturati) mostrano dualita dipolare? Il drift verso Poisson (POISSON_CONVERGENCE) e perdita di dualita dipolare?", "manuale": true, "porta": "sessione_interattiva", "condensato_ref": "A2,A10,F5", "condensato_motivo": "Discrimina due forme di det. A2 (confine) e la soglia. A10 (dipolo) e il tipo 1. F5 (frame) misura la struttura D-ND che e tipo 1." }, { "tipo": "scoperta_numerica", "id": "METRIC_TENSOR", "claim": "Il tensore metrico dei primi è g=(p/2)². Nel tempo ln(p), è de Sitter 1+1D. z=-8.8 curvatura vs z=+22.5 rapporti ΔΓ.", "intensità": 0.9, "nota": "Sessione interattiva 4 aprile. Verificato su 78K primi.", "manuale": true, "porta": "sessione_interattiva", "condensato_ref": null, "condensato_motivo": "Risultato numerico verificato, non-tautologico" }, { "tipo": "scoperta", "id": "TENSIONE_ENTITA", "claim": "La tensione non e un problema pratico - e un Entita. La tensione superflua crea latenza (tempo). Senza tensione superflua tutto e regolato da assiomi. Implicazione: le tensioni nel seme sono entita, non problemi da risolvere. Quelle superflue (det=+1) producono tempo/latenza.", "intensita": 0.85, "nota": "Input operatore 2026-04-10. Tocca: rapporto tensione/assioma. Operativamente: discriminare tensioni-entita (generative) da tensioni-superflue (dispersive) nel seme. Le 21 tensioni attuali - quante sono entita e quante latenza?", "manuale": true, "porta": "sessione_interattiva", "condensato_ref": "A5,A6", "condensato_motivo": "Il ciclo (A5) lavora con tensioni - ma se la tensione e entita, il ciclo non le risolve, le osserva. Lo zero mobile (A6) e la tensione senza latenza." }, { "tipo": "confine_inesplorato", "id": "G_POTENZIALE_NULLA", "claim": "G e il potenziale di tutto come nulla - permette il prima e il dopo. Ci muoviamo come trascendenza dimensionale gravitazionale. G nel tetraedro non e una teoria tra le altre - e il potenziale che le rende possibili.", "intensita": 0.85, "nota": "Input operatore 2026-04-10. Tocca: ruolo di G nel tetraedro (T,Q,G,E). La fonte video_lp0RgZ6kQF8 dice: tensore metrico dentro la forma simplettica. G non e accanto a T,Q,E - e sotto. Consecutio: nei dati Lab, i ponti TxG e ExG hanno struttura diversa dai ponti TxQ?", "manuale": true, "porta": "sessione_interattiva", "condensato_ref": "A7,A10", "condensato_motivo": "A7 (singolarita come operatore) e G come potenziale. A10 (dipolo) opera sul piano che G rende possibile." }, { "tipo": "confine_inesplorato", "id": "BOUNDARY", "claim": "8 domini GUE, 5 Poisson — il confine è il terzo incluso operativo", "intensità": 0.8, "nota": "Il segnale non-triviale è DOVE la scissione cambia natura, non che converge a φ", "condensato_ref": "A9", "condensato_motivo": "Overlap termini con A9 (5 termini)", "porta": "condensato" }, { "tipo": "confine_inesplorato", "id": "PIANO_PRIMARIO_DUE_ASSIOMI", "claim": "I piani importanti sono il primario e i due assiomi che lo determinano nelle zone osservate. Non tutti gli assiomi operano ovunque - in ogni zona osservata, due assiomi determinano il piano primario.", "intensita": 0.8, "nota": "Input operatore 2026-04-10. Tocca: struttura locale degli assiomi. Consecutio: per ogni dominio Lab (primi, logistica, percolazione...) quali 2 assiomi del condensato sono operativi? Mappa assiomi x domini = grafo della realta locale.", "manuale": true, "porta": "sessione_interattiva", "condensato_ref": "A9,A14", "condensato_motivo": "A9 (terzo incluso) opera CON il piano. A14 (cascata) propaga - ma propaga cosa, se solo 2 assiomi sono attivi per zona?" }, { "tipo": "simmetria_sospetta", "id": "META", "claim": "Tutti i 11 test passano — verifica che non stiamo testando solo tautologie", "intensità": 0.5, "nota": "La convergenza a φ è triviale (controprove). I test stanno verificando contenuto o struttura?", "condensato_ref": "A4,A12,C2", "porta": "verify_assertions_META_ALL_PASS", "condensato_motivo": "Ricorrente (3x in 2 giorni) e fuori dalla mappa" }, { "id": "OBSERVABLE_REGISTRY", "tipo": "vincolo", "claim": "Ogni script che usa observables canonici (SR, SR2, L1, L2, triple_var) deve importare la definizione da tools/observables_registry.py. Varianti devono usare nomi distinti (SR_local_rigidity, triple_var_normalized) — niente shadowing del nome canonico. Ogni report deve dichiarare 'observables_registry: VERSION' nel header.", "intensita": 1.0, "porta": "infrastructure", "manuale": true, "condensato_ref": "A14,A8", "origine": "cristallizzato 06/05 dalla consecutio del cycle 20260506_0625 (autopoietico self-finding)", "added_at": "2026-05-06T07:03:58.213606+00:00" } ], "potenziale_bloccato": [], "varianza": [ "Tensioni risolte: {'PIANO_PRIMARIO_DUE_ASSIOMI', 'TENSIONE_ENTITA', 'TRASCENDENZA_LIMITE', 'METRIC_TENSOR', 'DUALITA_DIPOLARE_VS_ILLUSORIA', 'G_POTENZIALE_NULLA'}" ], "filtro": { "promosse": 8, "filtrate": 0, "condensato_claims": 30 }, "direzione": "Esplorare il confine: 8 domini GUE, 5 Poisson — il confine è il terzo incluso operativo", "verifica": { "pass": 11, "fail": 0, "skip": 0, "total": 11 }, "fonti_consumate": 0, "fonti_esterne": [ { "id": "video_lp0RgZ6kQF8", "title": "Equivalence between geometrical structures and entropy", "type": "video_digest", "keywords": [ "geometry", "entropy", "symplectic form", "statistical mechanics", "quantum", "thermodynamics", "inner product", "Born rule", "metric tensor", "electromagnetic tensor" ], "content": "La geometria degli stati (classici e quantistici) e l'entropia sono la stessa struttura — invertibili. La forma simplettica conta le configurazioni. Il tensore metrico dello spaziotempo appare dentro la forma simplettica estesa. Il tensore elettromagnetico pure. Statistical mechanics non è costruita sopra alla meccanica — è la stessa cosa.", "teorie": [ "T", "Q", "G", "E" ], "ponti_potenziali": [ { "coppia": "TxQ", "ponte": "forma simplettica = entropia (invertibili)", "nota": "geometry is entropy and entropy is geometry" }, { "coppia": "TxG", "ponte": "tensore metrico dentro la forma simplettica estesa", "nota": "geometria spaziotempo = geometria degli stati in posizione×velocità" }, { "coppia": "ExT", "ponte": "tensore EM dentro la forma simplettica", "nota": "il campo EM conta stati in configurazione posizione×tempo" } ], "timestamp": "2026-04-02T08:23:13.991997" }, { "id": "video_sDlZ-aY9GN4", "title": "Moving charges produce magnetic fields - Einstein relativity", "type": "video_digest", "keywords": [ "magnetic field", "electric field", "length contraction", "time dilation", "Coulomb", "Lorentz", "reference frame", "electromagnetic" ], "content": "Il campo magnetico non esiste come entità separata — è il campo elettrico visto da un altro frame. La contrazione di Lorentz trasforma neutralità in carica. Due elettroni in movimento si separano più lentamente del previsto non per forza magnetica ma per dilatazione temporale. E e B sono manifestazioni dello stesso campo elettromagnetico. La relatività unifica.", "teorie": [ "E", "R" ], "ponti_potenziali": [ { "coppia": "ExR", "ponte": "cambio di frame — E e B sono lo stesso campo", "nota": "il 'ponte' non è l'onda EM, è il cambio di osservatore" } ], "timestamp": "2026-04-02T08:23:13.992016" }, { "id": "video_OwDWOtFNsKQ", "title": "Thermodynamic Computing: Better than Quantum? | Guillaume Verdon (Extropic)", "type": "video_digest", "keywords": [ "thermodynamic computing", "noise", "entropy", "extropic", "fluctuations", "information theory", "probability", "Boltzmann" ], "content": "Computing termodinamico: sfruttare le fluttuazioni termiche invece di combatterle. Il rumore non è nemico — è risorsa. Extropic costruisce hardware che usa l'entropia come motore computazionale. Connessione profonda tra termodinamica, teoria dell'informazione e probabilità.", "teorie": [ "T", "Q" ], "ponti_potenziali": [ { "coppia": "TxQ", "ponte": "noise come risorsa computazionale — fluttuazioni termiche = calcolo", "nota": "il vuoto quantistico (pieno di fluttuazioni) è il computer" } ], "timestamp": "2026-04-02T08:23:13.992019" }, { "id": "video_j0wJBEZdwLs", exec /bin/bash -lc "sed -n '280,430p' tools/exp_perturbation_dimensionality_audit.py" in /opt/MM_D-ND succeeded in 0ms: def replicate_summary(results: list[dict]) -> dict: ranks = np.array([r["pca"]["effective_rank"] for r in results], dtype=float) pc2 = np.array([r["pca"]["explained_variance"][1] if len(r["pca"]["explained_variance"]) > 1 else 0.0 for r in results]) cos = np.array([ r["pca"]["centroid_cosine"].get("adjacent_swap_vs_large_gap_only", 0.0) for r in results ]) return { "n_replicates": len(results), "effective_rank_mean": float(np.mean(ranks)), "effective_rank_std": float(np.std(ranks, ddof=1)) if len(ranks) > 1 else 0.0, "effective_rank_min": float(np.min(ranks)), "effective_rank_max": float(np.max(ranks)), "pc2_mean": float(np.mean(pc2)), "pc2_std": float(np.std(pc2, ddof=1)) if len(pc2) > 1 else 0.0, "adjacent_vs_large_cosine_mean": float(np.mean(cos)), "adjacent_vs_large_cosine_std": float(np.std(cos, ddof=1)) if len(cos) > 1 else 0.0, } def run(args: argparse.Namespace) -> dict: root_rng = np.random.default_rng(args.seed) alphas = [float(x) for x in np.linspace(args.alpha_min, args.alpha_max, args.n_alpha)] prime = prime_gaps(args.n_prime_gaps) fixed_domains = { "primes": prime, "prime_shuffle_control": root_rng.permutation(prime), "poisson": root_rng.exponential(1.0, size=args.n_prime_gaps), } output = { "experiment": "perturbation_dimensionality_audit", "question": "Is the GUE second perturbation axis stable across independent ensembles and sample-size controls?", "params": vars(args), "alphas": alphas, "observables": OBS_NAMES, "perturbations": PERT_NAMES, "fixed_domains": {}, "gue_replicates": [], "gue_summary": {}, "gue_short_replicates": [], "gue_short_summary": {}, } print("fixed domains") print(f"{'domain':<22} {'N':>7} {'rank':>7} {'PC2':>7} {'cos(adj,large)':>15}") for name, gaps in fixed_domains.items(): rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) res = analyze(name, gaps.astype(float), alphas, args.n_trials, args.n_baseline, rng) output["fixed_domains"][name] = res pc2 = res["pca"]["explained_variance"][1] cos = res["pca"]["centroid_cosine"]["adjacent_swap_vs_large_gap_only"] print(f"{name:<22} {len(gaps):>7} {res['pca']['effective_rank']:>7.3f} {pc2:>7.3f} {cos:>15.3f}") print("\nGUE independent replicates") for i in range(args.n_gue_replicates): rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) gaps = gue_spacings(args.gue_matrix_size, args.gue_matrices, rng) res = analyze(f"gue_rep_{i}", gaps, alphas, args.n_trials, args.n_baseline, rng) output["gue_replicates"].append(res) pc2 = res["pca"]["explained_variance"][1] cos = res["pca"]["centroid_cosine"]["adjacent_swap_vs_large_gap_only"] print(f"gue_rep_{i:<14} {len(gaps):>7} {res['pca']['effective_rank']:>7.3f} {pc2:>7.3f} {cos:>15.3f}") for i in range(args.n_gue_replicates): rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) gaps = gue_spacings(args.gue_short_matrix_size, 1, rng) res = analyze(f"gue_short_rep_{i}", gaps, alphas, args.n_trials, args.n_baseline, rng) output["gue_short_replicates"].append(res) output["gue_summary"] = replicate_summary(output["gue_replicates"]) output["gue_short_summary"] = replicate_summary(output["gue_short_replicates"]) print("\nsummary") for label, summary in [("gue", output["gue_summary"]), ("gue_short", output["gue_short_summary"])]: print( f"{label:<10} rank={summary['effective_rank_mean']:.3f}+/-{summary['effective_rank_std']:.3f} " f"PC2={summary['pc2_mean']:.3f}+/-{summary['pc2_std']:.3f} " f"cos={summary['adjacent_vs_large_cosine_mean']:.3f}+/-{summary['adjacent_vs_large_cosine_std']:.3f}" ) out_path = Path(args.out) out_path.parent.mkdir(parents=True, exist_ok=True) with out_path.open("w") as f: json.dump(output, f, indent=2) print(f"\nsaved {out_path}") return output def main() -> None: global OBSERVABLE_SET parser = argparse.ArgumentParser() parser.add_argument("--n-prime-gaps", type=int, default=12000) parser.add_argument("--gue-matrix-size", type=int, default=180) parser.add_argument("--gue-matrices", type=int, default=16) parser.add_argument("--gue-short-matrix-size", type=int, default=42) parser.add_argument("--n-gue-replicates", type=int, default=6) parser.add_argument("--n-alpha", type=int, default=5) parser.add_argument("--alpha-min", type=float, default=0.1) parser.add_argument("--alpha-max", type=float, default=0.9) parser.add_argument("--n-trials", type=int, default=10) parser.add_argument("--n-baseline", type=int, default=24) parser.add_argument("--seed", type=int, default=20260506) parser.add_argument("--observable-set", choices=["rank_audit", "scale_0330"], default="rank_audit") parser.add_argument("--out", default="tools/data/perturbation_dimensionality_audit.json") args = parser.parse_args() OBSERVABLE_SET = args.observable_set run(args) if __name__ == "__main__": main() codex Sto per aggiungere `tools/exp_perturbation_rank_size_curve.py`. La misura separa due cose che prima erano mescolate: rank apparente delle perturbazioni e forza dei denominatori original-vs-shuffle degli osservabili canonici. apply patch patch: completed /opt/MM_D-ND/tools/exp_perturbation_rank_size_curve.py diff --git a/tools/exp_perturbation_rank_size_curve.py b/tools/exp_perturbation_rank_size_curve.py new file mode 100644 index 0000000000000000000000000000000000000000..032847ca3d82ec3233ec2b41ad230dfe394d2791 --- /dev/null +++ b/tools/exp_perturbation_rank_size_curve.py @@ -0,0 +1,387 @@ +#!/usr/bin/env python3 +""" +exp_perturbation_rank_size_curve.py + +Reusable META audit for perturbation dimensionality. + +The 2026-05-06 06:25 cycle restricted the claim "GUE has a second +perturbation axis" to sample size, generator, and observable definitions. +This tool measures the size curve directly, using the canonical observable +registry and explicit original-vs-shuffle denominator diagnostics. + +The report owns interpretation. This script only measures: +- effective rank and PC2 across scale-selective perturbation profiles; +- original-vs-shuffle z-score per observable; +- whether apparent rank co-occurs with weak retention denominators. +""" + +from __future__ import annotations + +import argparse +import json +from pathlib import Path + +import numpy as np + +from observables_registry import ( + OBSERVABLES_CANONICAL, + OBSERVABLES_REGISTRY_VERSION, + compute_canonical, +) + + +OBS_NAMES = list(OBSERVABLES_CANONICAL.keys()) +PERT_NAMES = ["adjacent_swap", "block_shuffle", "large_gap_only", "uniform"] + + +def prime_gaps(n_gaps: int) -> np.ndarray: + """Return the first n_gaps prime gaps using a compact numpy sieve.""" + limit = max(100, int(n_gaps * (np.log(n_gaps + 10) + np.log(np.log(n_gaps + 10)) + 5))) + while True: + sieve = np.ones(limit + 1, dtype=bool) + sieve[:2] = False + for p in range(2, int(limit**0.5) + 1): + if sieve[p]: + sieve[p * p : limit + 1 : p] = False + primes = np.flatnonzero(sieve) + if len(primes) >= n_gaps + 1: + return np.diff(primes[: n_gaps + 1]).astype(float) + limit *= 2 + + +def gue_spacings(matrix_size: int, min_spacings: int, rng: np.random.Generator) -> np.ndarray: + """Generate unfolded GUE spacings by concatenating independent matrices.""" + parts = [] + edge = max(2, matrix_size // 10) + while sum(len(x) for x in parts) < min_spacings: + real = rng.standard_normal((matrix_size, matrix_size)) + imag = rng.standard_normal((matrix_size, matrix_size)) + h = real + 1j * imag + h = (h + h.conj().T) / (2.0 * np.sqrt(matrix_size)) + eigs = np.sort(np.linalg.eigvalsh(h).real) + bulk = eigs[edge:-edge] + gaps = np.diff(bulk) + mean = float(np.mean(gaps)) + if mean > 1e-15: + parts.append(gaps / mean) + return np.concatenate(parts)[:min_spacings].astype(float) + + +def perturb_adjacent_swap(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + idx = np.arange(0, len(out) - 1, 2) + chosen = idx[rng.random(len(idx)) < alpha] + tmp = out[chosen].copy() + out[chosen] = out[chosen + 1] + out[chosen + 1] = tmp + return out + + +def perturb_block_shuffle(gaps: np.ndarray, alpha: float, rng: np.random.Generator, block_size: int = 64) -> np.ndarray: + out = gaps.copy() + n_blocks = len(out) // block_size + if n_blocks <= 0: + return rng.permutation(out) if alpha > 0 else out + k = int(round(alpha * n_blocks)) + if k <= 0: + return out + for block in rng.choice(n_blocks, size=min(k, n_blocks), replace=False): + start = block * block_size + end = min(start + block_size, len(out)) + rng.shuffle(out[start:end]) + return out + + +def perturb_large_gap_only(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + idx = np.flatnonzero(out > np.median(out)) + k = int(round(alpha * len(idx))) + if k < 2: + return out + chosen = rng.choice(idx, size=min(k, len(idx)), replace=False) + vals = out[chosen].copy() + rng.shuffle(vals) + out[chosen] = vals + return out + + +def perturb_uniform(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + k = int(round(alpha * len(out))) + if k < 2: + return out + chosen = rng.choice(len(out), size=min(k, len(out)), replace=False) + vals = out[chosen].copy() + rng.shuffle(vals) + out[chosen] = vals + return out + + +PERTURB = { + "adjacent_swap": perturb_adjacent_swap, + "block_shuffle": perturb_block_shuffle, + "large_gap_only": perturb_large_gap_only, + "uniform": perturb_uniform, +} + + +def pca_summary(vectors: list[list[float]], labels: list[str]) -> dict: + matrix = np.array(vectors, dtype=float) + matrix = matrix - np.mean(matrix, axis=0, keepdims=True) + if min(matrix.shape) == 0 or np.max(np.abs(matrix)) <= 1e-15: + return { + "explained_variance": [], + "effective_rank": 0.0, + "centroid_cosine": {}, + "pc2": 0.0, + } + _, singular, _ = np.linalg.svd(matrix, full_matrices=False) + energy = singular * singular + if float(np.sum(energy)) <= 1e-15: + explained = np.zeros_like(energy) + effective_rank = 0.0 + else: + explained = energy / np.sum(energy) + pos = explained[explained > 1e-15] + effective_rank = float(np.exp(-np.sum(pos * np.log(pos)))) + + centroids = {} + for name in PERT_NAMES: + vals = np.array([v for v, label in zip(vectors, labels) if label == name], dtype=float) + if len(vals): + centroids[name] = np.mean(vals, axis=0) + + cosine = {} + for i, a_name in enumerate(PERT_NAMES): + for b_name in PERT_NAMES[i + 1 :]: + if a_name not in centroids or b_name not in centroids: + continue + a = centroids[a_name] + b = centroids[b_name] + denom = np.linalg.norm(a) * np.linalg.norm(b) + cosine[f"{a_name}_vs_{b_name}"] = float(np.dot(a, b) / denom) if denom > 1e-15 else 0.0 + + return { + "explained_variance": [float(x) for x in explained], + "effective_rank": effective_rank, + "centroid_cosine": cosine, + "pc2": float(explained[1]) if len(explained) > 1 else 0.0, + } + + +def analyze_sequence( + gaps: np.ndarray, + alphas: list[float], + n_trials: int, + n_baseline: int, + z_min: float, + rng: np.random.Generator, +) -> dict: + original = compute_canonical(gaps) + baseline_vals = {obs: [] for obs in OBS_NAMES} + for _ in range(n_baseline): + obs = compute_canonical(rng.permutation(gaps)) + for name in OBS_NAMES: + baseline_vals[name].append(obs[name]) + + baseline = {} + z = {} + denom = {} + for name in OBS_NAMES: + vals = np.array(baseline_vals[name], dtype=float) + mean = float(np.mean(vals)) + std = float(np.std(vals, ddof=1)) if len(vals) > 1 else 0.0 + baseline[name] = {"mean": mean, "std": std} + denom[name] = float(original[name] - mean) + z[name] = float(denom[name] / std) if std > 1e-15 else 0.0 + + stable_obs = [name for name in OBS_NAMES if abs(z[name]) >= z_min] + all_vectors = [] + screened_vectors = [] + labels = [] + profiles = [] + + for pert_name in PERT_NAMES: + for alpha in alphas: + trial_vals = {obs: [] for obs in OBS_NAMES} + for _ in range(n_trials): + perturbed = PERTURB[pert_name](gaps, alpha, rng) + obs = compute_canonical(perturbed) + for name in OBS_NAMES: + trial_vals[name].append(obs[name]) + means = {name: float(np.mean(trial_vals[name])) for name in OBS_NAMES} + retention = {} + for name in OBS_NAMES: + retention[name] = ( + float((means[name] - baseline[name]["mean"]) / denom[name]) + if abs(denom[name]) > 1e-12 + else 0.0 + ) + all_vector = [retention[name] for name in OBS_NAMES] + screened_vector = [retention[name] for name in stable_obs] + all_vectors.append(all_vector) + if len(stable_obs) >= 2: + screened_vectors.append(screened_vector) + labels.append(pert_name) + profiles.append( + { + "perturbation": pert_name, + "alpha": float(alpha), + "retention": retention, + "retention_vector": all_vector, + } + ) + + all_pca = pca_summary(all_vectors, labels) + screened_pca = pca_summary(screened_vectors, labels) if len(stable_obs) >= 2 else None + + return { + "n_gaps": int(len(gaps)), + "original": original, + "full_shuffle_baseline": baseline, + "denominator": denom, + "original_vs_shuffle_z": z, + "stable_observables": stable_obs, + "weak_observable_count": int(len(OBS_NAMES) - len(stable_obs)), + "profiles": profiles, + "pca_all_observables": all_pca, + "pca_stable_observables": screened_pca, + } + + +def summarize_replicates(items: list[dict]) -> dict: + def arr(path: tuple[str, ...]) -> np.ndarray: + vals = [] + for item in items: + x = item + for key in path: + if x is None: + break + x = x.get(key) + if isinstance(x, (int, float)): + vals.append(float(x)) + return np.array(vals, dtype=float) + + rank = arr(("pca_all_observables", "effective_rank")) + pc2 = arr(("pca_all_observables", "pc2")) + weak = np.array([item["weak_observable_count"] for item in items], dtype=float) + stable_rank = arr(("pca_stable_observables", "effective_rank")) + cos = arr(("pca_all_observables", "centroid_cosine", "adjacent_swap_vs_large_gap_only")) + + out = { + "n_replicates": len(items), + "rank_mean": float(np.mean(rank)) if len(rank) else 0.0, + "rank_std": float(np.std(rank, ddof=1)) if len(rank) > 1 else 0.0, + "pc2_mean": float(np.mean(pc2)) if len(pc2) else 0.0, + "pc2_std": float(np.std(pc2, ddof=1)) if len(pc2) > 1 else 0.0, + "weak_observable_count_mean": float(np.mean(weak)) if len(weak) else 0.0, + "stable_rank_mean": float(np.mean(stable_rank)) if len(stable_rank) else None, + "stable_rank_std": float(np.std(stable_rank, ddof=1)) if len(stable_rank) > 1 else 0.0, + "adjacent_vs_large_cosine_mean": float(np.mean(cos)) if len(cos) else 0.0, + "adjacent_vs_large_cosine_std": float(np.std(cos, ddof=1)) if len(cos) > 1 else 0.0, + } + if len(rank) > 1 and np.std(weak) > 1e-15 and np.std(rank) > 1e-15: + out["rank_vs_weak_count_corr"] = float(np.corrcoef(rank, weak)[0, 1]) + else: + out["rank_vs_weak_count_corr"] = 0.0 + return out + + +def build_prime_windows(max_n: int, n_reps: int) -> list[np.ndarray]: + total = max_n * n_reps + max_n + gaps = prime_gaps(total) + max_start = len(gaps) - max_n + starts = np.linspace(0, max_start, n_reps, dtype=int) + return [gaps[start : start + max_n].astype(float) for start in starts] + + +def run(args: argparse.Namespace) -> dict: + root_rng = np.random.default_rng(args.seed) + sizes = [int(x) for x in args.sizes.split(",") if x.strip()] + max_n = max(sizes) + alphas = [float(x) for x in np.linspace(args.alpha_min, args.alpha_max, args.n_alpha)] + + output = { + "experiment": "perturbation_rank_size_curve", + "question": "Does perturbation effective-rank stabilize with sample size under canonical observables?", + "observables_registry": OBSERVABLES_REGISTRY_VERSION, + "observables_used": OBS_NAMES, + "perturbations": PERT_NAMES, + "params": vars(args), + "alphas": alphas, + "domains": {}, + "summary": {}, + } + + prime_windows = build_prime_windows(max_n, args.n_replicates) + + print(f"observables_registry={OBSERVABLES_REGISTRY_VERSION}") + print(f"{'domain':<22} {'N':>6} {'rank':>7} {'PC2':>7} {'weak':>5} {'stable_rank':>11}") + + domain_builders = { + "primes_windows": lambda rep_rng, rep_i: prime_windows[rep_i], + "prime_shuffle_control": lambda rep_rng, rep_i: rep_rng.permutation(prime_windows[rep_i]), + "poisson": lambda rep_rng, rep_i: rep_rng.exponential(1.0, size=max_n), + "gue": lambda rep_rng, rep_i: gue_spacings(args.gue_matrix_size, max_n, rep_rng), + } + + for domain_name, builder in domain_builders.items(): + output["domains"][domain_name] = {} + output["summary"][domain_name] = {} + bases = [] + for rep_i in range(args.n_replicates): + rep_rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) + bases.append(builder(rep_rng, rep_i)) + + for n in sizes: + rows = [] + for rep_i, base in enumerate(bases): + rep_rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) + res = analyze_sequence( + base[:n], + alphas=alphas, + n_trials=args.n_trials, + n_baseline=args.n_baseline, + z_min=args.z_min, + rng=rep_rng, + ) + res["replicate"] = rep_i + rows.append(res) + summary = summarize_replicates(rows) + output["domains"][domain_name][str(n)] = rows + output["summary"][domain_name][str(n)] = summary + stable_rank = summary["stable_rank_mean"] + stable_text = f"{stable_rank:>11.3f}" if stable_rank is not None else f"{'NA':>11}" + print( + f"{domain_name:<22} {n:>6} {summary['rank_mean']:>7.3f} " + f"{summary['pc2_mean']:>7.3f} {summary['weak_observable_count_mean']:>5.2f} " + f"{stable_text}" + ) + + out_path = Path(args.out) + out_path.parent.mkdir(parents=True, exist_ok=True) + with out_path.open("w") as f: + json.dump(output, f, indent=2) + print(f"saved {out_path}") + return output + + +def main() -> None: + parser = argparse.ArgumentParser() + parser.add_argument("--sizes", default="128,256,512,1024,2048") + parser.add_argument("--n-replicates", type=int, default=8) + parser.add_argument("--gue-matrix-size", type=int, default=180) + parser.add_argument("--n-alpha", type=int, default=5) + parser.add_argument("--alpha-min", type=float, default=0.1) + parser.add_argument("--alpha-max", type=float, default=0.9) + parser.add_argument("--n-trials", type=int, default=8) + parser.add_argument("--n-baseline", type=int, default=16) + parser.add_argument("--z-min", type=float, default=2.0) + parser.add_argument("--seed", type=int, default=20260506) + parser.add_argument("--out", default="tools/data/perturbation_rank_size_curve.json") + args = parser.parse_args() + run(args) + + +if __name__ == "__main__": + main() diff --git a/tools/exp_perturbation_rank_size_curve.py b/tools/exp_perturbation_rank_size_curve.py new file mode 100644 index 0000000000000000000000000000000000000000..032847ca3d82ec3233ec2b41ad230dfe394d2791 --- /dev/null +++ b/tools/exp_perturbation_rank_size_curve.py @@ -0,0 +1,387 @@ +#!/usr/bin/env python3 +""" +exp_perturbation_rank_size_curve.py + +Reusable META audit for perturbation dimensionality. + +The 2026-05-06 06:25 cycle restricted the claim "GUE has a second +perturbation axis" to sample size, generator, and observable definitions. +This tool measures the size curve directly, using the canonical observable +registry and explicit original-vs-shuffle denominator diagnostics. + +The report owns interpretation. This script only measures: +- effective rank and PC2 across scale-selective perturbation profiles; +- original-vs-shuffle z-score per observable; +- whether apparent rank co-occurs with weak retention denominators. +""" + +from __future__ import annotations + +import argparse +import json +from pathlib import Path + +import numpy as np + +from observables_registry import ( + OBSERVABLES_CANONICAL, + OBSERVABLES_REGISTRY_VERSION, + compute_canonical, +) + + +OBS_NAMES = list(OBSERVABLES_CANONICAL.keys()) +PERT_NAMES = ["adjacent_swap", "block_shuffle", "large_gap_only", "uniform"] + + +def prime_gaps(n_gaps: int) -> np.ndarray: + """Return the first n_gaps prime gaps using a compact numpy sieve.""" + limit = max(100, int(n_gaps * (np.log(n_gaps + 10) + np.log(np.log(n_gaps + 10)) + 5))) + while True: + sieve = np.ones(limit + 1, dtype=bool) + sieve[:2] = False + for p in range(2, int(limit**0.5) + 1): + if sieve[p]: + sieve[p * p : limit + 1 : p] = False + primes = np.flatnonzero(sieve) + if len(primes) >= n_gaps + 1: + return np.diff(primes[: n_gaps + 1]).astype(float) + limit *= 2 + + +def gue_spacings(matrix_size: int, min_spacings: int, rng: np.random.Generator) -> np.ndarray: + """Generate unfolded GUE spacings by concatenating independent matrices.""" + parts = [] + edge = max(2, matrix_size // 10) + while sum(len(x) for x in parts) < min_spacings: + real = rng.standard_normal((matrix_size, matrix_size)) + imag = rng.standard_normal((matrix_size, matrix_size)) + h = real + 1j * imag + h = (h + h.conj().T) / (2.0 * np.sqrt(matrix_size)) + eigs = np.sort(np.linalg.eigvalsh(h).real) + bulk = eigs[edge:-edge] + gaps = np.diff(bulk) + mean = float(np.mean(gaps)) + if mean > 1e-15: + parts.append(gaps / mean) + return np.concatenate(parts)[:min_spacings].astype(float) + + +def perturb_adjacent_swap(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + idx = np.arange(0, len(out) - 1, 2) + chosen = idx[rng.random(len(idx)) < alpha] + tmp = out[chosen].copy() + out[chosen] = out[chosen + 1] + out[chosen + 1] = tmp + return out + + +def perturb_block_shuffle(gaps: np.ndarray, alpha: float, rng: np.random.Generator, block_size: int = 64) -> np.ndarray: + out = gaps.copy() + n_blocks = len(out) // block_size + if n_blocks <= 0: + return rng.permutation(out) if alpha > 0 else out + k = int(round(alpha * n_blocks)) + if k <= 0: + return out + for block in rng.choice(n_blocks, size=min(k, n_blocks), replace=False): + start = block * block_size + end = min(start + block_size, len(out)) + rng.shuffle(out[start:end]) + return out + + +def perturb_large_gap_only(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + idx = np.flatnonzero(out > np.median(out)) + k = int(round(alpha * len(idx))) + if k < 2: + return out + chosen = rng.choice(idx, size=min(k, len(idx)), replace=False) + vals = out[chosen].copy() + rng.shuffle(vals) + out[chosen] = vals + return out + + +def perturb_uniform(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + k = int(round(alpha * len(out))) + if k < 2: + return out + chosen = rng.choice(len(out), size=min(k, len(out)), replace=False) + vals = out[chosen].copy() + rng.shuffle(vals) + out[chosen] = vals + return out + + +PERTURB = { + "adjacent_swap": perturb_adjacent_swap, + "block_shuffle": perturb_block_shuffle, + "large_gap_only": perturb_large_gap_only, + "uniform": perturb_uniform, +} + + +def pca_summary(vectors: list[list[float]], labels: list[str]) -> dict: + matrix = np.array(vectors, dtype=float) + matrix = matrix - np.mean(matrix, axis=0, keepdims=True) + if min(matrix.shape) == 0 or np.max(np.abs(matrix)) <= 1e-15: + return { + "explained_variance": [], + "effective_rank": 0.0, + "centroid_cosine": {}, + "pc2": 0.0, + } + _, singular, _ = np.linalg.svd(matrix, full_matrices=False) + energy = singular * singular + if float(np.sum(energy)) <= 1e-15: + explained = np.zeros_like(energy) + effective_rank = 0.0 + else: + explained = energy / np.sum(energy) + pos = explained[explained > 1e-15] + effective_rank = float(np.exp(-np.sum(pos * np.log(pos)))) + + centroids = {} + for name in PERT_NAMES: + vals = np.array([v for v, label in zip(vectors, labels) if label == name], dtype=float) + if len(vals): + centroids[name] = np.mean(vals, axis=0) + + cosine = {} + for i, a_name in enumerate(PERT_NAMES): + for b_name in PERT_NAMES[i + 1 :]: + if a_name not in centroids or b_name not in centroids: + continue + a = centroids[a_name] + b = centroids[b_name] + denom = np.linalg.norm(a) * np.linalg.norm(b) + cosine[f"{a_name}_vs_{b_name}"] = float(np.dot(a, b) / denom) if denom > 1e-15 else 0.0 + + return { + "explained_variance": [float(x) for x in explained], + "effective_rank": effective_rank, + "centroid_cosine": cosine, + "pc2": float(explained[1]) if len(explained) > 1 else 0.0, + } + + +def analyze_sequence( + gaps: np.ndarray, + alphas: list[float], + n_trials: int, + n_baseline: int, + z_min: float, + rng: np.random.Generator, +) -> dict: + original = compute_canonical(gaps) + baseline_vals = {obs: [] for obs in OBS_NAMES} + for _ in range(n_baseline): + obs = compute_canonical(rng.permutation(gaps)) + for name in OBS_NAMES: + baseline_vals[name].append(obs[name]) + + baseline = {} + z = {} + denom = {} + for name in OBS_NAMES: + vals = np.array(baseline_vals[name], dtype=float) + mean = float(np.mean(vals)) + std = float(np.std(vals, ddof=1)) if len(vals) > 1 else 0.0 + baseline[name] = {"mean": mean, "std": std} + denom[name] = float(original[name] - mean) + z[name] = float(denom[name] / std) if std > 1e-15 else 0.0 + + stable_obs = [name for name in OBS_NAMES if abs(z[name]) >= z_min] + all_vectors = [] + screened_vectors = [] + labels = [] + profiles = [] + + for pert_name in PERT_NAMES: + for alpha in alphas: + trial_vals = {obs: [] for obs in OBS_NAMES} + for _ in range(n_trials): + perturbed = PERTURB[pert_name](gaps, alpha, rng) + obs = compute_canonical(perturbed) + for name in OBS_NAMES: + trial_vals[name].append(obs[name]) + means = {name: float(np.mean(trial_vals[name])) for name in OBS_NAMES} + retention = {} + for name in OBS_NAMES: + retention[name] = ( + float((means[name] - baseline[name]["mean"]) / denom[name]) + if abs(denom[name]) > 1e-12 + else 0.0 + ) + all_vector = [retention[name] for name in OBS_NAMES] + screened_vector = [retention[name] for name in stable_obs] + all_vectors.append(all_vector) + if len(stable_obs) >= 2: + screened_vectors.append(screened_vector) + labels.append(pert_name) + profiles.append( + { + "perturbation": pert_name, + "alpha": float(alpha), + "retention": retention, + "retention_vector": all_vector, + } + ) + + all_pca = pca_summary(all_vectors, labels) + screened_pca = pca_summary(screened_vectors, labels) if len(stable_obs) >= 2 else None + + return { + "n_gaps": int(len(gaps)), + "original": original, + "full_shuffle_baseline": baseline, + "denominator": denom, + "original_vs_shuffle_z": z, + "stable_observables": stable_obs, + "weak_observable_count": int(len(OBS_NAMES) - len(stable_obs)), + "profiles": profiles, + "pca_all_observables": all_pca, + "pca_stable_observables": screened_pca, + } + + +def summarize_replicates(items: list[dict]) -> dict: + def arr(path: tuple[str, ...]) -> np.ndarray: + vals = [] + for item in items: + x = item + for key in path: + if x is None: + break + x = x.get(key) + if isinstance(x, (int, float)): + vals.append(float(x)) + return np.array(vals, dtype=float) + + rank = arr(("pca_all_observables", "effective_rank")) + pc2 = arr(("pca_all_observables", "pc2")) + weak = np.array([item["weak_observable_count"] for item in items], dtype=float) + stable_rank = arr(("pca_stable_observables", "effective_rank")) + cos = arr(("pca_all_observables", "centroid_cosine", "adjacent_swap_vs_large_gap_only")) + + out = { + "n_replicates": len(items), + "rank_mean": float(np.mean(rank)) if len(rank) else 0.0, + "rank_std": float(np.std(rank, ddof=1)) if len(rank) > 1 else 0.0, + "pc2_mean": float(np.mean(pc2)) if len(pc2) else 0.0, + "pc2_std": float(np.std(pc2, ddof=1)) if len(pc2) > 1 else 0.0, + "weak_observable_count_mean": float(np.mean(weak)) if len(weak) else 0.0, + "stable_rank_mean": float(np.mean(stable_rank)) if len(stable_rank) else None, + "stable_rank_std": float(np.std(stable_rank, ddof=1)) if len(stable_rank) > 1 else 0.0, + "adjacent_vs_large_cosine_mean": float(np.mean(cos)) if len(cos) else 0.0, + "adjacent_vs_large_cosine_std": float(np.std(cos, ddof=1)) if len(cos) > 1 else 0.0, + } + if len(rank) > 1 and np.std(weak) > 1e-15 and np.std(rank) > 1e-15: + out["rank_vs_weak_count_corr"] = float(np.corrcoef(rank, weak)[0, 1]) + else: + out["rank_vs_weak_count_corr"] = 0.0 + return out + + +def build_prime_windows(max_n: int, n_reps: int) -> list[np.ndarray]: + total = max_n * n_reps + max_n + gaps = prime_gaps(total) + max_start = len(gaps) - max_n + starts = np.linspace(0, max_start, n_reps, dtype=int) + return [gaps[start : start + max_n].astype(float) for start in starts] + + +def run(args: argparse.Namespace) -> dict: + root_rng = np.random.default_rng(args.seed) + sizes = [int(x) for x in args.sizes.split(",") if x.strip()] + max_n = max(sizes) + alphas = [float(x) for x in np.linspace(args.alpha_min, args.alpha_max, args.n_alpha)] + + output = { + "experiment": "perturbation_rank_size_curve", + "question": "Does perturbation effective-rank stabilize with sample size under canonical observables?", + "observables_registry": OBSERVABLES_REGISTRY_VERSION, + "observables_used": OBS_NAMES, + "perturbations": PERT_NAMES, + "params": vars(args), + "alphas": alphas, + "domains": {}, + "summary": {}, + } + + prime_windows = build_prime_windows(max_n, args.n_replicates) + + print(f"observables_registry={OBSERVABLES_REGISTRY_VERSION}") + print(f"{'domain':<22} {'N':>6} {'rank':>7} {'PC2':>7} {'weak':>5} {'stable_rank':>11}") + + domain_builders = { + "primes_windows": lambda rep_rng, rep_i: prime_windows[rep_i], + "prime_shuffle_control": lambda rep_rng, rep_i: rep_rng.permutation(prime_windows[rep_i]), + "poisson": lambda rep_rng, rep_i: rep_rng.exponential(1.0, size=max_n), + "gue": lambda rep_rng, rep_i: gue_spacings(args.gue_matrix_size, max_n, rep_rng), + } + + for domain_name, builder in domain_builders.items(): + output["domains"][domain_name] = {} + output["summary"][domain_name] = {} + bases = [] + for rep_i in range(args.n_replicates): + rep_rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) + bases.append(builder(rep_rng, rep_i)) + + for n in sizes: + rows = [] + for rep_i, base in enumerate(bases): + rep_rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) + res = analyze_sequence( + base[:n], + alphas=alphas, + n_trials=args.n_trials, + n_baseline=args.n_baseline, + z_min=args.z_min, + rng=rep_rng, + ) + res["replicate"] = rep_i + rows.append(res) + summary = summarize_replicates(rows) + output["domains"][domain_name][str(n)] = rows + output["summary"][domain_name][str(n)] = summary + stable_rank = summary["stable_rank_mean"] + stable_text = f"{stable_rank:>11.3f}" if stable_rank is not None else f"{'NA':>11}" + print( + f"{domain_name:<22} {n:>6} {summary['rank_mean']:>7.3f} " + f"{summary['pc2_mean']:>7.3f} {summary['weak_observable_count_mean']:>5.2f} " + f"{stable_text}" + ) + + out_path = Path(args.out) + out_path.parent.mkdir(parents=True, exist_ok=True) + with out_path.open("w") as f: + json.dump(output, f, indent=2) + print(f"saved {out_path}") + return output + + +def main() -> None: + parser = argparse.ArgumentParser() + parser.add_argument("--sizes", default="128,256,512,1024,2048") + parser.add_argument("--n-replicates", type=int, default=8) + parser.add_argument("--gue-matrix-size", type=int, default=180) + parser.add_argument("--n-alpha", type=int, default=5) + parser.add_argument("--alpha-min", type=float, default=0.1) + parser.add_argument("--alpha-max", type=float, default=0.9) + parser.add_argument("--n-trials", type=int, default=8) + parser.add_argument("--n-baseline", type=int, default=16) + parser.add_argument("--z-min", type=float, default=2.0) + parser.add_argument("--seed", type=int, default=20260506) + parser.add_argument("--out", default="tools/data/perturbation_rank_size_curve.json") + args = parser.parse_args() + run(args) + + +if __name__ == "__main__": + main() codex Lo strumento è in posto. Faccio prima un run piccolo di smoke test per intercettare errori, poi il run completo con parametri abbastanza leggeri da restare riproducibile nel ciclo. exec /bin/bash -lc 'python tools/exp_perturbation_rank_size_curve.py --sizes 64,128 --n-replicates 2 --n-trials 2 --n-baseline 4 --gue-matrix-size 60 --out /tmp/perturbation_rank_size_curve_smoke.json' in /opt/MM_D-ND diff --git a/tools/exp_perturbation_rank_size_curve.py b/tools/exp_perturbation_rank_size_curve.py new file mode 100644 index 0000000000000000000000000000000000000000..032847ca3d82ec3233ec2b41ad230dfe394d2791 --- /dev/null +++ b/tools/exp_perturbation_rank_size_curve.py @@ -0,0 +1,387 @@ +#!/usr/bin/env python3 +""" +exp_perturbation_rank_size_curve.py + +Reusable META audit for perturbation dimensionality. + +The 2026-05-06 06:25 cycle restricted the claim "GUE has a second +perturbation axis" to sample size, generator, and observable definitions. +This tool measures the size curve directly, using the canonical observable +registry and explicit original-vs-shuffle denominator diagnostics. + +The report owns interpretation. This script only measures: +- effective rank and PC2 across scale-selective perturbation profiles; +- original-vs-shuffle z-score per observable; +- whether apparent rank co-occurs with weak retention denominators. +""" + +from __future__ import annotations + +import argparse +import json +from pathlib import Path + +import numpy as np + +from observables_registry import ( + OBSERVABLES_CANONICAL, + OBSERVABLES_REGISTRY_VERSION, + compute_canonical, +) + + +OBS_NAMES = list(OBSERVABLES_CANONICAL.keys()) +PERT_NAMES = ["adjacent_swap", "block_shuffle", "large_gap_only", "uniform"] + + +def prime_gaps(n_gaps: int) -> np.ndarray: + """Return the first n_gaps prime gaps using a compact numpy sieve.""" + limit = max(100, int(n_gaps * (np.log(n_gaps + 10) + np.log(np.log(n_gaps + 10)) + 5))) + while True: + sieve = np.ones(limit + 1, dtype=bool) + sieve[:2] = False + for p in range(2, int(limit**0.5) + 1): + if sieve[p]: + sieve[p * p : limit + 1 : p] = False + primes = np.flatnonzero(sieve) + if len(primes) >= n_gaps + 1: + return np.diff(primes[: n_gaps + 1]).astype(float) + limit *= 2 + + +def gue_spacings(matrix_size: int, min_spacings: int, rng: np.random.Generator) -> np.ndarray: + """Generate unfolded GUE spacings by concatenating independent matrices.""" + parts = [] + edge = max(2, matrix_size // 10) + while sum(len(x) for x in parts) < min_spacings: + real = rng.standard_normal((matrix_size, matrix_size)) + imag = rng.standard_normal((matrix_size, matrix_size)) + h = real + 1j * imag + h = (h + h.conj().T) / (2.0 * np.sqrt(matrix_size)) + eigs = np.sort(np.linalg.eigvalsh(h).real) + bulk = eigs[edge:-edge] + gaps = np.diff(bulk) + mean = float(np.mean(gaps)) + if mean > 1e-15: + parts.append(gaps / mean) + return np.concatenate(parts)[:min_spacings].astype(float) + + +def perturb_adjacent_swap(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + idx = np.arange(0, len(out) - 1, 2) + chosen = idx[rng.random(len(idx)) < alpha] + tmp = out[chosen].copy() + out[chosen] = out[chosen + 1] + out[chosen + 1] = tmp + return out + + +def perturb_block_shuffle(gaps: np.ndarray, alpha: float, rng: np.random.Generator, block_size: int = 64) -> np.ndarray: + out = gaps.copy() + n_blocks = len(out) // block_size + if n_blocks <= 0: + return rng.permutation(out) if alpha > 0 else out + k = int(round(alpha * n_blocks)) + if k <= 0: + return out + for block in rng.choice(n_blocks, size=min(k, n_blocks), replace=False): + start = block * block_size + end = min(start + block_size, len(out)) + rng.shuffle(out[start:end]) + return out + + +def perturb_large_gap_only(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + idx = np.flatnonzero(out > np.median(out)) + k = int(round(alpha * len(idx))) + if k < 2: + return out + chosen = rng.choice(idx, size=min(k, len(idx)), replace=False) + vals = out[chosen].copy() + rng.shuffle(vals) + out[chosen] = vals + return out + + +def perturb_uniform(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + k = int(round(alpha * len(out))) + if k < 2: + return out + chosen = rng.choice(len(out), size=min(k, len(out)), replace=False) + vals = out[chosen].copy() + rng.shuffle(vals) + out[chosen] = vals + return out + + +PERTURB = { + "adjacent_swap": perturb_adjacent_swap, + "block_shuffle": perturb_block_shuffle, + "large_gap_only": perturb_large_gap_only, + "uniform": perturb_uniform, +} + + +def pca_summary(vectors: list[list[float]], labels: list[str]) -> dict: + matrix = np.array(vectors, dtype=float) + matrix = matrix - np.mean(matrix, axis=0, keepdims=True) + if min(matrix.shape) == 0 or np.max(np.abs(matrix)) <= 1e-15: + return { + "explained_variance": [], + "effective_rank": 0.0, + "centroid_cosine": {}, + "pc2": 0.0, + } + _, singular, _ = np.linalg.svd(matrix, full_matrices=False) + energy = singular * singular + if float(np.sum(energy)) <= 1e-15: + explained = np.zeros_like(energy) + effective_rank = 0.0 + else: + explained = energy / np.sum(energy) + pos = explained[explained > 1e-15] + effective_rank = float(np.exp(-np.sum(pos * np.log(pos)))) + + centroids = {} + for name in PERT_NAMES: + vals = np.array([v for v, label in zip(vectors, labels) if label == name], dtype=float) + if len(vals): + centroids[name] = np.mean(vals, axis=0) + + cosine = {} + for i, a_name in enumerate(PERT_NAMES): + for b_name in PERT_NAMES[i + 1 :]: + if a_name not in centroids or b_name not in centroids: + continue + a = centroids[a_name] + b = centroids[b_name] + denom = np.linalg.norm(a) * np.linalg.norm(b) + cosine[f"{a_name}_vs_{b_name}"] = float(np.dot(a, b) / denom) if denom > 1e-15 else 0.0 + + return { + "explained_variance": [float(x) for x in explained], + "effective_rank": effective_rank, + "centroid_cosine": cosine, + "pc2": float(explained[1]) if len(explained) > 1 else 0.0, + } + + +def analyze_sequence( + gaps: np.ndarray, + alphas: list[float], + n_trials: int, + n_baseline: int, + z_min: float, + rng: np.random.Generator, +) -> dict: + original = compute_canonical(gaps) + baseline_vals = {obs: [] for obs in OBS_NAMES} + for _ in range(n_baseline): + obs = compute_canonical(rng.permutation(gaps)) + for name in OBS_NAMES: + baseline_vals[name].append(obs[name]) + + baseline = {} + z = {} + denom = {} + for name in OBS_NAMES: + vals = np.array(baseline_vals[name], dtype=float) + mean = float(np.mean(vals)) + std = float(np.std(vals, ddof=1)) if len(vals) > 1 else 0.0 + baseline[name] = {"mean": mean, "std": std} + denom[name] = float(original[name] - mean) + z[name] = float(denom[name] / std) if std > 1e-15 else 0.0 + + stable_obs = [name for name in OBS_NAMES if abs(z[name]) >= z_min] + all_vectors = [] + screened_vectors = [] + labels = [] + profiles = [] + + for pert_name in PERT_NAMES: + for alpha in alphas: + trial_vals = {obs: [] for obs in OBS_NAMES} + for _ in range(n_trials): + perturbed = PERTURB[pert_name](gaps, alpha, rng) + obs = compute_canonical(perturbed) + for name in OBS_NAMES: + trial_vals[name].append(obs[name]) + means = {name: float(np.mean(trial_vals[name])) for name in OBS_NAMES} + retention = {} + for name in OBS_NAMES: + retention[name] = ( + float((means[name] - baseline[name]["mean"]) / denom[name]) + if abs(denom[name]) > 1e-12 + else 0.0 + ) + all_vector = [retention[name] for name in OBS_NAMES] + screened_vector = [retention[name] for name in stable_obs] + all_vectors.append(all_vector) + if len(stable_obs) >= 2: + screened_vectors.append(screened_vector) + labels.append(pert_name) + profiles.append( + { + "perturbation": pert_name, + "alpha": float(alpha), + "retention": retention, + "retention_vector": all_vector, + } + ) + + all_pca = pca_summary(all_vectors, labels) + screened_pca = pca_summary(screened_vectors, labels) if len(stable_obs) >= 2 else None + + return { + "n_gaps": int(len(gaps)), + "original": original, + "full_shuffle_baseline": baseline, + "denominator": denom, + "original_vs_shuffle_z": z, + "stable_observables": stable_obs, + "weak_observable_count": int(len(OBS_NAMES) - len(stable_obs)), + "profiles": profiles, + "pca_all_observables": all_pca, + "pca_stable_observables": screened_pca, + } + + +def summarize_replicates(items: list[dict]) -> dict: + def arr(path: tuple[str, ...]) -> np.ndarray: + vals = [] + for item in items: + x = item + for key in path: + if x is None: + break + x = x.get(key) + if isinstance(x, (int, float)): + vals.append(float(x)) + return np.array(vals, dtype=float) + + rank = arr(("pca_all_observables", "effective_rank")) + pc2 = arr(("pca_all_observables", "pc2")) + weak = np.array([item["weak_observable_count"] for item in items], dtype=float) + stable_rank = arr(("pca_stable_observables", "effective_rank")) + cos = arr(("pca_all_observables", "centroid_cosine", "adjacent_swap_vs_large_gap_only")) + + out = { + "n_replicates": len(items), + "rank_mean": float(np.mean(rank)) if len(rank) else 0.0, + "rank_std": float(np.std(rank, ddof=1)) if len(rank) > 1 else 0.0, + "pc2_mean": float(np.mean(pc2)) if len(pc2) else 0.0, + "pc2_std": float(np.std(pc2, ddof=1)) if len(pc2) > 1 else 0.0, + "weak_observable_count_mean": float(np.mean(weak)) if len(weak) else 0.0, + "stable_rank_mean": float(np.mean(stable_rank)) if len(stable_rank) else None, + "stable_rank_std": float(np.std(stable_rank, ddof=1)) if len(stable_rank) > 1 else 0.0, + "adjacent_vs_large_cosine_mean": float(np.mean(cos)) if len(cos) else 0.0, + "adjacent_vs_large_cosine_std": float(np.std(cos, ddof=1)) if len(cos) > 1 else 0.0, + } + if len(rank) > 1 and np.std(weak) > 1e-15 and np.std(rank) > 1e-15: + out["rank_vs_weak_count_corr"] = float(np.corrcoef(rank, weak)[0, 1]) + else: + out["rank_vs_weak_count_corr"] = 0.0 + return out + + +def build_prime_windows(max_n: int, n_reps: int) -> list[np.ndarray]: + total = max_n * n_reps + max_n + gaps = prime_gaps(total) + max_start = len(gaps) - max_n + starts = np.linspace(0, max_start, n_reps, dtype=int) + return [gaps[start : start + max_n].astype(float) for start in starts] + + +def run(args: argparse.Namespace) -> dict: + root_rng = np.random.default_rng(args.seed) + sizes = [int(x) for x in args.sizes.split(",") if x.strip()] + max_n = max(sizes) + alphas = [float(x) for x in np.linspace(args.alpha_min, args.alpha_max, args.n_alpha)] + + output = { + "experiment": "perturbation_rank_size_curve", + "question": "Does perturbation effective-rank stabilize with sample size under canonical observables?", + "observables_registry": OBSERVABLES_REGISTRY_VERSION, + "observables_used": OBS_NAMES, + "perturbations": PERT_NAMES, + "params": vars(args), + "alphas": alphas, + "domains": {}, + "summary": {}, + } + + prime_windows = build_prime_windows(max_n, args.n_replicates) + + print(f"observables_registry={OBSERVABLES_REGISTRY_VERSION}") + print(f"{'domain':<22} {'N':>6} {'rank':>7} {'PC2':>7} {'weak':>5} {'stable_rank':>11}") + + domain_builders = { + "primes_windows": lambda rep_rng, rep_i: prime_windows[rep_i], + "prime_shuffle_control": lambda rep_rng, rep_i: rep_rng.permutation(prime_windows[rep_i]), + "poisson": lambda rep_rng, rep_i: rep_rng.exponential(1.0, size=max_n), + "gue": lambda rep_rng, rep_i: gue_spacings(args.gue_matrix_size, max_n, rep_rng), + } + + for domain_name, builder in domain_builders.items(): + output["domains"][domain_name] = {} + output["summary"][domain_name] = {} + bases = [] + for rep_i in range(args.n_replicates): + rep_rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) + bases.append(builder(rep_rng, rep_i)) + + for n in sizes: + rows = [] + for rep_i, base in enumerate(bases): + rep_rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) + res = analyze_sequence( + base[:n], + alphas=alphas, + n_trials=args.n_trials, + n_baseline=args.n_baseline, + z_min=args.z_min, + rng=rep_rng, + ) + res["replicate"] = rep_i + rows.append(res) + summary = summarize_replicates(rows) + output["domains"][domain_name][str(n)] = rows + output["summary"][domain_name][str(n)] = summary + stable_rank = summary["stable_rank_mean"] + stable_text = f"{stable_rank:>11.3f}" if stable_rank is not None else f"{'NA':>11}" + print( + f"{domain_name:<22} {n:>6} {summary['rank_mean']:>7.3f} " + f"{summary['pc2_mean']:>7.3f} {summary['weak_observable_count_mean']:>5.2f} " + f"{stable_text}" + ) + + out_path = Path(args.out) + out_path.parent.mkdir(parents=True, exist_ok=True) + with out_path.open("w") as f: + json.dump(output, f, indent=2) + print(f"saved {out_path}") + return output + + +def main() -> None: + parser = argparse.ArgumentParser() + parser.add_argument("--sizes", default="128,256,512,1024,2048") + parser.add_argument("--n-replicates", type=int, default=8) + parser.add_argument("--gue-matrix-size", type=int, default=180) + parser.add_argument("--n-alpha", type=int, default=5) + parser.add_argument("--alpha-min", type=float, default=0.1) + parser.add_argument("--alpha-max", type=float, default=0.9) + parser.add_argument("--n-trials", type=int, default=8) + parser.add_argument("--n-baseline", type=int, default=16) + parser.add_argument("--z-min", type=float, default=2.0) + parser.add_argument("--seed", type=int, default=20260506) + parser.add_argument("--out", default="tools/data/perturbation_rank_size_curve.json") + args = parser.parse_args() + run(args) + + +if __name__ == "__main__": + main() succeeded in 298ms: observables_registry=1.0.0-2026-05-06 domain N rank PC2 weak stable_rank primes_windows 64 2.294 0.188 4.50 NA primes_windows 128 1.634 0.124 4.00 NA prime_shuffle_control 64 2.132 0.190 3.50 1.787 prime_shuffle_control 128 2.182 0.313 4.50 NA poisson 64 2.422 0.216 3.50 1.559 poisson 128 1.293 0.035 5.00 NA gue 64 1.324 0.060 2.50 1.096 gue 128 1.038 0.004 2.50 1.189 saved /tmp/perturbation_rank_size_curve_smoke.json exec /bin/bash -lc 'python tools/exp_perturbation_rank_size_curve.py --sizes 128,256,512,1024,2048 --n-replicates 8 --n-trials 8 --n-baseline 16 --gue-matrix-size 180 --out tools/data/perturbation_rank_size_curve.json' in /opt/MM_D-ND diff --git a/tools/exp_perturbation_rank_size_curve.py b/tools/exp_perturbation_rank_size_curve.py new file mode 100644 index 0000000000000000000000000000000000000000..032847ca3d82ec3233ec2b41ad230dfe394d2791 --- /dev/null +++ b/tools/exp_perturbation_rank_size_curve.py @@ -0,0 +1,387 @@ +#!/usr/bin/env python3 +""" +exp_perturbation_rank_size_curve.py + +Reusable META audit for perturbation dimensionality. + +The 2026-05-06 06:25 cycle restricted the claim "GUE has a second +perturbation axis" to sample size, generator, and observable definitions. +This tool measures the size curve directly, using the canonical observable +registry and explicit original-vs-shuffle denominator diagnostics. + +The report owns interpretation. This script only measures: +- effective rank and PC2 across scale-selective perturbation profiles; +- original-vs-shuffle z-score per observable; +- whether apparent rank co-occurs with weak retention denominators. +""" + +from __future__ import annotations + +import argparse +import json +from pathlib import Path + +import numpy as np + +from observables_registry import ( + OBSERVABLES_CANONICAL, + OBSERVABLES_REGISTRY_VERSION, + compute_canonical, +) + + +OBS_NAMES = list(OBSERVABLES_CANONICAL.keys()) +PERT_NAMES = ["adjacent_swap", "block_shuffle", "large_gap_only", "uniform"] + + +def prime_gaps(n_gaps: int) -> np.ndarray: + """Return the first n_gaps prime gaps using a compact numpy sieve.""" + limit = max(100, int(n_gaps * (np.log(n_gaps + 10) + np.log(np.log(n_gaps + 10)) + 5))) + while True: + sieve = np.ones(limit + 1, dtype=bool) + sieve[:2] = False + for p in range(2, int(limit**0.5) + 1): + if sieve[p]: + sieve[p * p : limit + 1 : p] = False + primes = np.flatnonzero(sieve) + if len(primes) >= n_gaps + 1: + return np.diff(primes[: n_gaps + 1]).astype(float) + limit *= 2 + + +def gue_spacings(matrix_size: int, min_spacings: int, rng: np.random.Generator) -> np.ndarray: + """Generate unfolded GUE spacings by concatenating independent matrices.""" + parts = [] + edge = max(2, matrix_size // 10) + while sum(len(x) for x in parts) < min_spacings: + real = rng.standard_normal((matrix_size, matrix_size)) + imag = rng.standard_normal((matrix_size, matrix_size)) + h = real + 1j * imag + h = (h + h.conj().T) / (2.0 * np.sqrt(matrix_size)) + eigs = np.sort(np.linalg.eigvalsh(h).real) + bulk = eigs[edge:-edge] + gaps = np.diff(bulk) + mean = float(np.mean(gaps)) + if mean > 1e-15: + parts.append(gaps / mean) + return np.concatenate(parts)[:min_spacings].astype(float) + + +def perturb_adjacent_swap(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + idx = np.arange(0, len(out) - 1, 2) + chosen = idx[rng.random(len(idx)) < alpha] + tmp = out[chosen].copy() + out[chosen] = out[chosen + 1] + out[chosen + 1] = tmp + return out + + +def perturb_block_shuffle(gaps: np.ndarray, alpha: float, rng: np.random.Generator, block_size: int = 64) -> np.ndarray: + out = gaps.copy() + n_blocks = len(out) // block_size + if n_blocks <= 0: + return rng.permutation(out) if alpha > 0 else out + k = int(round(alpha * n_blocks)) + if k <= 0: + return out + for block in rng.choice(n_blocks, size=min(k, n_blocks), replace=False): + start = block * block_size + end = min(start + block_size, len(out)) + rng.shuffle(out[start:end]) + return out + + +def perturb_large_gap_only(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + idx = np.flatnonzero(out > np.median(out)) + k = int(round(alpha * len(idx))) + if k < 2: + return out + chosen = rng.choice(idx, size=min(k, len(idx)), replace=False) + vals = out[chosen].copy() + rng.shuffle(vals) + out[chosen] = vals + return out + + +def perturb_uniform(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + k = int(round(alpha * len(out))) + if k < 2: + return out + chosen = rng.choice(len(out), size=min(k, len(out)), replace=False) + vals = out[chosen].copy() + rng.shuffle(vals) + out[chosen] = vals + return out + + +PERTURB = { + "adjacent_swap": perturb_adjacent_swap, + "block_shuffle": perturb_block_shuffle, + "large_gap_only": perturb_large_gap_only, + "uniform": perturb_uniform, +} + + +def pca_summary(vectors: list[list[float]], labels: list[str]) -> dict: + matrix = np.array(vectors, dtype=float) + matrix = matrix - np.mean(matrix, axis=0, keepdims=True) + if min(matrix.shape) == 0 or np.max(np.abs(matrix)) <= 1e-15: + return { + "explained_variance": [], + "effective_rank": 0.0, + "centroid_cosine": {}, + "pc2": 0.0, + } + _, singular, _ = np.linalg.svd(matrix, full_matrices=False) + energy = singular * singular + if float(np.sum(energy)) <= 1e-15: + explained = np.zeros_like(energy) + effective_rank = 0.0 + else: + explained = energy / np.sum(energy) + pos = explained[explained > 1e-15] + effective_rank = float(np.exp(-np.sum(pos * np.log(pos)))) + + centroids = {} + for name in PERT_NAMES: + vals = np.array([v for v, label in zip(vectors, labels) if label == name], dtype=float) + if len(vals): + centroids[name] = np.mean(vals, axis=0) + + cosine = {} + for i, a_name in enumerate(PERT_NAMES): + for b_name in PERT_NAMES[i + 1 :]: + if a_name not in centroids or b_name not in centroids: + continue + a = centroids[a_name] + b = centroids[b_name] + denom = np.linalg.norm(a) * np.linalg.norm(b) + cosine[f"{a_name}_vs_{b_name}"] = float(np.dot(a, b) / denom) if denom > 1e-15 else 0.0 + + return { + "explained_variance": [float(x) for x in explained], + "effective_rank": effective_rank, + "centroid_cosine": cosine, + "pc2": float(explained[1]) if len(explained) > 1 else 0.0, + } + + +def analyze_sequence( + gaps: np.ndarray, + alphas: list[float], + n_trials: int, + n_baseline: int, + z_min: float, + rng: np.random.Generator, +) -> dict: + original = compute_canonical(gaps) + baseline_vals = {obs: [] for obs in OBS_NAMES} + for _ in range(n_baseline): + obs = compute_canonical(rng.permutation(gaps)) + for name in OBS_NAMES: + baseline_vals[name].append(obs[name]) + + baseline = {} + z = {} + denom = {} + for name in OBS_NAMES: + vals = np.array(baseline_vals[name], dtype=float) + mean = float(np.mean(vals)) + std = float(np.std(vals, ddof=1)) if len(vals) > 1 else 0.0 + baseline[name] = {"mean": mean, "std": std} + denom[name] = float(original[name] - mean) + z[name] = float(denom[name] / std) if std > 1e-15 else 0.0 + + stable_obs = [name for name in OBS_NAMES if abs(z[name]) >= z_min] + all_vectors = [] + screened_vectors = [] + labels = [] + profiles = [] + + for pert_name in PERT_NAMES: + for alpha in alphas: + trial_vals = {obs: [] for obs in OBS_NAMES} + for _ in range(n_trials): + perturbed = PERTURB[pert_name](gaps, alpha, rng) + obs = compute_canonical(perturbed) + for name in OBS_NAMES: + trial_vals[name].append(obs[name]) + means = {name: float(np.mean(trial_vals[name])) for name in OBS_NAMES} + retention = {} + for name in OBS_NAMES: + retention[name] = ( + float((means[name] - baseline[name]["mean"]) / denom[name]) + if abs(denom[name]) > 1e-12 + else 0.0 + ) + all_vector = [retention[name] for name in OBS_NAMES] + screened_vector = [retention[name] for name in stable_obs] + all_vectors.append(all_vector) + if len(stable_obs) >= 2: + screened_vectors.append(screened_vector) + labels.append(pert_name) + profiles.append( + { + "perturbation": pert_name, + "alpha": float(alpha), + "retention": retention, + "retention_vector": all_vector, + } + ) + + all_pca = pca_summary(all_vectors, labels) + screened_pca = pca_summary(screened_vectors, labels) if len(stable_obs) >= 2 else None + + return { + "n_gaps": int(len(gaps)), + "original": original, + "full_shuffle_baseline": baseline, + "denominator": denom, + "original_vs_shuffle_z": z, + "stable_observables": stable_obs, + "weak_observable_count": int(len(OBS_NAMES) - len(stable_obs)), + "profiles": profiles, + "pca_all_observables": all_pca, + "pca_stable_observables": screened_pca, + } + + +def summarize_replicates(items: list[dict]) -> dict: + def arr(path: tuple[str, ...]) -> np.ndarray: + vals = [] + for item in items: + x = item + for key in path: + if x is None: + break + x = x.get(key) + if isinstance(x, (int, float)): + vals.append(float(x)) + return np.array(vals, dtype=float) + + rank = arr(("pca_all_observables", "effective_rank")) + pc2 = arr(("pca_all_observables", "pc2")) + weak = np.array([item["weak_observable_count"] for item in items], dtype=float) + stable_rank = arr(("pca_stable_observables", "effective_rank")) + cos = arr(("pca_all_observables", "centroid_cosine", "adjacent_swap_vs_large_gap_only")) + + out = { + "n_replicates": len(items), + "rank_mean": float(np.mean(rank)) if len(rank) else 0.0, + "rank_std": float(np.std(rank, ddof=1)) if len(rank) > 1 else 0.0, + "pc2_mean": float(np.mean(pc2)) if len(pc2) else 0.0, + "pc2_std": float(np.std(pc2, ddof=1)) if len(pc2) > 1 else 0.0, + "weak_observable_count_mean": float(np.mean(weak)) if len(weak) else 0.0, + "stable_rank_mean": float(np.mean(stable_rank)) if len(stable_rank) else None, + "stable_rank_std": float(np.std(stable_rank, ddof=1)) if len(stable_rank) > 1 else 0.0, + "adjacent_vs_large_cosine_mean": float(np.mean(cos)) if len(cos) else 0.0, + "adjacent_vs_large_cosine_std": float(np.std(cos, ddof=1)) if len(cos) > 1 else 0.0, + } + if len(rank) > 1 and np.std(weak) > 1e-15 and np.std(rank) > 1e-15: + out["rank_vs_weak_count_corr"] = float(np.corrcoef(rank, weak)[0, 1]) + else: + out["rank_vs_weak_count_corr"] = 0.0 + return out + + +def build_prime_windows(max_n: int, n_reps: int) -> list[np.ndarray]: + total = max_n * n_reps + max_n + gaps = prime_gaps(total) + max_start = len(gaps) - max_n + starts = np.linspace(0, max_start, n_reps, dtype=int) + return [gaps[start : start + max_n].astype(float) for start in starts] + + +def run(args: argparse.Namespace) -> dict: + root_rng = np.random.default_rng(args.seed) + sizes = [int(x) for x in args.sizes.split(",") if x.strip()] + max_n = max(sizes) + alphas = [float(x) for x in np.linspace(args.alpha_min, args.alpha_max, args.n_alpha)] + + output = { + "experiment": "perturbation_rank_size_curve", + "question": "Does perturbation effective-rank stabilize with sample size under canonical observables?", + "observables_registry": OBSERVABLES_REGISTRY_VERSION, + "observables_used": OBS_NAMES, + "perturbations": PERT_NAMES, + "params": vars(args), + "alphas": alphas, + "domains": {}, + "summary": {}, + } + + prime_windows = build_prime_windows(max_n, args.n_replicates) + + print(f"observables_registry={OBSERVABLES_REGISTRY_VERSION}") + print(f"{'domain':<22} {'N':>6} {'rank':>7} {'PC2':>7} {'weak':>5} {'stable_rank':>11}") + + domain_builders = { + "primes_windows": lambda rep_rng, rep_i: prime_windows[rep_i], + "prime_shuffle_control": lambda rep_rng, rep_i: rep_rng.permutation(prime_windows[rep_i]), + "poisson": lambda rep_rng, rep_i: rep_rng.exponential(1.0, size=max_n), + "gue": lambda rep_rng, rep_i: gue_spacings(args.gue_matrix_size, max_n, rep_rng), + } + + for domain_name, builder in domain_builders.items(): + output["domains"][domain_name] = {} + output["summary"][domain_name] = {} + bases = [] + for rep_i in range(args.n_replicates): + rep_rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) + bases.append(builder(rep_rng, rep_i)) + + for n in sizes: + rows = [] + for rep_i, base in enumerate(bases): + rep_rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) + res = analyze_sequence( + base[:n], + alphas=alphas, + n_trials=args.n_trials, + n_baseline=args.n_baseline, + z_min=args.z_min, + rng=rep_rng, + ) + res["replicate"] = rep_i + rows.append(res) + summary = summarize_replicates(rows) + output["domains"][domain_name][str(n)] = rows + output["summary"][domain_name][str(n)] = summary + stable_rank = summary["stable_rank_mean"] + stable_text = f"{stable_rank:>11.3f}" if stable_rank is not None else f"{'NA':>11}" + print( + f"{domain_name:<22} {n:>6} {summary['rank_mean']:>7.3f} " + f"{summary['pc2_mean']:>7.3f} {summary['weak_observable_count_mean']:>5.2f} " + f"{stable_text}" + ) + + out_path = Path(args.out) + out_path.parent.mkdir(parents=True, exist_ok=True) + with out_path.open("w") as f: + json.dump(output, f, indent=2) + print(f"saved {out_path}") + return output + + +def main() -> None: + parser = argparse.ArgumentParser() + parser.add_argument("--sizes", default="128,256,512,1024,2048") + parser.add_argument("--n-replicates", type=int, default=8) + parser.add_argument("--gue-matrix-size", type=int, default=180) + parser.add_argument("--n-alpha", type=int, default=5) + parser.add_argument("--alpha-min", type=float, default=0.1) + parser.add_argument("--alpha-max", type=float, default=0.9) + parser.add_argument("--n-trials", type=int, default=8) + parser.add_argument("--n-baseline", type=int, default=16) + parser.add_argument("--z-min", type=float, default=2.0) + parser.add_argument("--seed", type=int, default=20260506) + parser.add_argument("--out", default="tools/data/perturbation_rank_size_curve.json") + args = parser.parse_args() + run(args) + + +if __name__ == "__main__": + main() succeeded in 3396ms: observables_registry=1.0.0-2026-05-06 domain N rank PC2 weak stable_rank primes_windows 128 1.789 0.155 4.50 1.382 primes_windows 256 1.947 0.174 4.75 1.262 primes_windows 512 1.892 0.142 2.88 1.310 primes_windows 1024 1.679 0.117 1.62 1.415 primes_windows 2048 1.442 0.081 0.75 1.462 prime_shuffle_control 128 1.600 0.126 4.75 1.174 prime_shuffle_control 256 1.952 0.183 4.38 1.087 prime_shuffle_control 512 1.743 0.135 4.12 1.216 prime_shuffle_control 1024 1.711 0.119 4.38 NA prime_shuffle_control 2048 1.797 0.134 3.62 1.428 poisson 128 1.765 0.133 4.62 1.033 poisson 256 1.920 0.169 4.50 1.177 poisson 512 1.737 0.126 4.25 1.275 poisson 1024 1.881 0.176 4.50 1.055 poisson 2048 1.952 0.175 4.62 1.036 gue 128 1.703 0.126 2.38 1.226 gue 256 1.913 0.164 2.25 1.141 gue 512 1.542 0.111 1.88 1.162 gue 1024 1.551 0.105 1.88 1.157 gue 2048 1.234 0.046 2.00 1.111 saved tools/data/perturbation_rank_size_curve.json codex Il run completo è partito. Sta generando repliche GUE e finestre sui primi; mentre gira, il criterio di lettura sarà conservativo: nessun “asse” viene trattato come struttura se gli osservabili che lo alimentano hanno denominatori deboli rispetto allo shuffle. diff --git a/tools/exp_perturbation_rank_size_curve.py b/tools/exp_perturbation_rank_size_curve.py new file mode 100644 index 0000000000000000000000000000000000000000..032847ca3d82ec3233ec2b41ad230dfe394d2791 --- /dev/null +++ b/tools/exp_perturbation_rank_size_curve.py @@ -0,0 +1,387 @@ +#!/usr/bin/env python3 +""" +exp_perturbation_rank_size_curve.py + +Reusable META audit for perturbation dimensionality. + +The 2026-05-06 06:25 cycle restricted the claim "GUE has a second +perturbation axis" to sample size, generator, and observable definitions. +This tool measures the size curve directly, using the canonical observable +registry and explicit original-vs-shuffle denominator diagnostics. + +The report owns interpretation. This script only measures: +- effective rank and PC2 across scale-selective perturbation profiles; +- original-vs-shuffle z-score per observable; +- whether apparent rank co-occurs with weak retention denominators. +""" + +from __future__ import annotations + +import argparse +import json +from pathlib import Path + +import numpy as np + +from observables_registry import ( + OBSERVABLES_CANONICAL, + OBSERVABLES_REGISTRY_VERSION, + compute_canonical, +) + + +OBS_NAMES = list(OBSERVABLES_CANONICAL.keys()) +PERT_NAMES = ["adjacent_swap", "block_shuffle", "large_gap_only", "uniform"] + + +def prime_gaps(n_gaps: int) -> np.ndarray: + """Return the first n_gaps prime gaps using a compact numpy sieve.""" + limit = max(100, int(n_gaps * (np.log(n_gaps + 10) + np.log(np.log(n_gaps + 10)) + 5))) + while True: + sieve = np.ones(limit + 1, dtype=bool) + sieve[:2] = False + for p in range(2, int(limit**0.5) + 1): + if sieve[p]: + sieve[p * p : limit + 1 : p] = False + primes = np.flatnonzero(sieve) + if len(primes) >= n_gaps + 1: + return np.diff(primes[: n_gaps + 1]).astype(float) + limit *= 2 + + +def gue_spacings(matrix_size: int, min_spacings: int, rng: np.random.Generator) -> np.ndarray: + """Generate unfolded GUE spacings by concatenating independent matrices.""" + parts = [] + edge = max(2, matrix_size // 10) + while sum(len(x) for x in parts) < min_spacings: + real = rng.standard_normal((matrix_size, matrix_size)) + imag = rng.standard_normal((matrix_size, matrix_size)) + h = real + 1j * imag + h = (h + h.conj().T) / (2.0 * np.sqrt(matrix_size)) + eigs = np.sort(np.linalg.eigvalsh(h).real) + bulk = eigs[edge:-edge] + gaps = np.diff(bulk) + mean = float(np.mean(gaps)) + if mean > 1e-15: + parts.append(gaps / mean) + return np.concatenate(parts)[:min_spacings].astype(float) + + +def perturb_adjacent_swap(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + idx = np.arange(0, len(out) - 1, 2) + chosen = idx[rng.random(len(idx)) < alpha] + tmp = out[chosen].copy() + out[chosen] = out[chosen + 1] + out[chosen + 1] = tmp + return out + + +def perturb_block_shuffle(gaps: np.ndarray, alpha: float, rng: np.random.Generator, block_size: int = 64) -> np.ndarray: + out = gaps.copy() + n_blocks = len(out) // block_size + if n_blocks <= 0: + return rng.permutation(out) if alpha > 0 else out + k = int(round(alpha * n_blocks)) + if k <= 0: + return out + for block in rng.choice(n_blocks, size=min(k, n_blocks), replace=False): + start = block * block_size + end = min(start + block_size, len(out)) + rng.shuffle(out[start:end]) + return out + + +def perturb_large_gap_only(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + idx = np.flatnonzero(out > np.median(out)) + k = int(round(alpha * len(idx))) + if k < 2: + return out + chosen = rng.choice(idx, size=min(k, len(idx)), replace=False) + vals = out[chosen].copy() + rng.shuffle(vals) + out[chosen] = vals + return out + + +def perturb_uniform(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + k = int(round(alpha * len(out))) + if k < 2: + return out + chosen = rng.choice(len(out), size=min(k, len(out)), replace=False) + vals = out[chosen].copy() + rng.shuffle(vals) + out[chosen] = vals + return out + + +PERTURB = { + "adjacent_swap": perturb_adjacent_swap, + "block_shuffle": perturb_block_shuffle, + "large_gap_only": perturb_large_gap_only, + "uniform": perturb_uniform, +} + + +def pca_summary(vectors: list[list[float]], labels: list[str]) -> dict: + matrix = np.array(vectors, dtype=float) + matrix = matrix - np.mean(matrix, axis=0, keepdims=True) + if min(matrix.shape) == 0 or np.max(np.abs(matrix)) <= 1e-15: + return { + "explained_variance": [], + "effective_rank": 0.0, + "centroid_cosine": {}, + "pc2": 0.0, + } + _, singular, _ = np.linalg.svd(matrix, full_matrices=False) + energy = singular * singular + if float(np.sum(energy)) <= 1e-15: + explained = np.zeros_like(energy) + effective_rank = 0.0 + else: + explained = energy / np.sum(energy) + pos = explained[explained > 1e-15] + effective_rank = float(np.exp(-np.sum(pos * np.log(pos)))) + + centroids = {} + for name in PERT_NAMES: + vals = np.array([v for v, label in zip(vectors, labels) if label == name], dtype=float) + if len(vals): + centroids[name] = np.mean(vals, axis=0) + + cosine = {} + for i, a_name in enumerate(PERT_NAMES): + for b_name in PERT_NAMES[i + 1 :]: + if a_name not in centroids or b_name not in centroids: + continue + a = centroids[a_name] + b = centroids[b_name] + denom = np.linalg.norm(a) * np.linalg.norm(b) + cosine[f"{a_name}_vs_{b_name}"] = float(np.dot(a, b) / denom) if denom > 1e-15 else 0.0 + + return { + "explained_variance": [float(x) for x in explained], + "effective_rank": effective_rank, + "centroid_cosine": cosine, + "pc2": float(explained[1]) if len(explained) > 1 else 0.0, + } + + +def analyze_sequence( + gaps: np.ndarray, + alphas: list[float], + n_trials: int, + n_baseline: int, + z_min: float, + rng: np.random.Generator, +) -> dict: + original = compute_canonical(gaps) + baseline_vals = {obs: [] for obs in OBS_NAMES} + for _ in range(n_baseline): + obs = compute_canonical(rng.permutation(gaps)) + for name in OBS_NAMES: + baseline_vals[name].append(obs[name]) + + baseline = {} + z = {} + denom = {} + for name in OBS_NAMES: + vals = np.array(baseline_vals[name], dtype=float) + mean = float(np.mean(vals)) + std = float(np.std(vals, ddof=1)) if len(vals) > 1 else 0.0 + baseline[name] = {"mean": mean, "std": std} + denom[name] = float(original[name] - mean) + z[name] = float(denom[name] / std) if std > 1e-15 else 0.0 + + stable_obs = [name for name in OBS_NAMES if abs(z[name]) >= z_min] + all_vectors = [] + screened_vectors = [] + labels = [] + profiles = [] + + for pert_name in PERT_NAMES: + for alpha in alphas: + trial_vals = {obs: [] for obs in OBS_NAMES} + for _ in range(n_trials): + perturbed = PERTURB[pert_name](gaps, alpha, rng) + obs = compute_canonical(perturbed) + for name in OBS_NAMES: + trial_vals[name].append(obs[name]) + means = {name: float(np.mean(trial_vals[name])) for name in OBS_NAMES} + retention = {} + for name in OBS_NAMES: + retention[name] = ( + float((means[name] - baseline[name]["mean"]) / denom[name]) + if abs(denom[name]) > 1e-12 + else 0.0 + ) + all_vector = [retention[name] for name in OBS_NAMES] + screened_vector = [retention[name] for name in stable_obs] + all_vectors.append(all_vector) + if len(stable_obs) >= 2: + screened_vectors.append(screened_vector) + labels.append(pert_name) + profiles.append( + { + "perturbation": pert_name, + "alpha": float(alpha), + "retention": retention, + "retention_vector": all_vector, + } + ) + + all_pca = pca_summary(all_vectors, labels) + screened_pca = pca_summary(screened_vectors, labels) if len(stable_obs) >= 2 else None + + return { + "n_gaps": int(len(gaps)), + "original": original, + "full_shuffle_baseline": baseline, + "denominator": denom, + "original_vs_shuffle_z": z, + "stable_observables": stable_obs, + "weak_observable_count": int(len(OBS_NAMES) - len(stable_obs)), + "profiles": profiles, + "pca_all_observables": all_pca, + "pca_stable_observables": screened_pca, + } + + +def summarize_replicates(items: list[dict]) -> dict: + def arr(path: tuple[str, ...]) -> np.ndarray: + vals = [] + for item in items: + x = item + for key in path: + if x is None: + break + x = x.get(key) + if isinstance(x, (int, float)): + vals.append(float(x)) + return np.array(vals, dtype=float) + + rank = arr(("pca_all_observables", "effective_rank")) + pc2 = arr(("pca_all_observables", "pc2")) + weak = np.array([item["weak_observable_count"] for item in items], dtype=float) + stable_rank = arr(("pca_stable_observables", "effective_rank")) + cos = arr(("pca_all_observables", "centroid_cosine", "adjacent_swap_vs_large_gap_only")) + + out = { + "n_replicates": len(items), + "rank_mean": float(np.mean(rank)) if len(rank) else 0.0, + "rank_std": float(np.std(rank, ddof=1)) if len(rank) > 1 else 0.0, + "pc2_mean": float(np.mean(pc2)) if len(pc2) else 0.0, + "pc2_std": float(np.std(pc2, ddof=1)) if len(pc2) > 1 else 0.0, + "weak_observable_count_mean": float(np.mean(weak)) if len(weak) else 0.0, + "stable_rank_mean": float(np.mean(stable_rank)) if len(stable_rank) else None, + "stable_rank_std": float(np.std(stable_rank, ddof=1)) if len(stable_rank) > 1 else 0.0, + "adjacent_vs_large_cosine_mean": float(np.mean(cos)) if len(cos) else 0.0, + "adjacent_vs_large_cosine_std": float(np.std(cos, ddof=1)) if len(cos) > 1 else 0.0, + } + if len(rank) > 1 and np.std(weak) > 1e-15 and np.std(rank) > 1e-15: + out["rank_vs_weak_count_corr"] = float(np.corrcoef(rank, weak)[0, 1]) + else: + out["rank_vs_weak_count_corr"] = 0.0 + return out + + +def build_prime_windows(max_n: int, n_reps: int) -> list[np.ndarray]: + total = max_n * n_reps + max_n + gaps = prime_gaps(total) + max_start = len(gaps) - max_n + starts = np.linspace(0, max_start, n_reps, dtype=int) + return [gaps[start : start + max_n].astype(float) for start in starts] + + +def run(args: argparse.Namespace) -> dict: + root_rng = np.random.default_rng(args.seed) + sizes = [int(x) for x in args.sizes.split(",") if x.strip()] + max_n = max(sizes) + alphas = [float(x) for x in np.linspace(args.alpha_min, args.alpha_max, args.n_alpha)] + + output = { + "experiment": "perturbation_rank_size_curve", + "question": "Does perturbation effective-rank stabilize with sample size under canonical observables?", + "observables_registry": OBSERVABLES_REGISTRY_VERSION, + "observables_used": OBS_NAMES, + "perturbations": PERT_NAMES, + "params": vars(args), + "alphas": alphas, + "domains": {}, + "summary": {}, + } + + prime_windows = build_prime_windows(max_n, args.n_replicates) + + print(f"observables_registry={OBSERVABLES_REGISTRY_VERSION}") + print(f"{'domain':<22} {'N':>6} {'rank':>7} {'PC2':>7} {'weak':>5} {'stable_rank':>11}") + + domain_builders = { + "primes_windows": lambda rep_rng, rep_i: prime_windows[rep_i], + "prime_shuffle_control": lambda rep_rng, rep_i: rep_rng.permutation(prime_windows[rep_i]), + "poisson": lambda rep_rng, rep_i: rep_rng.exponential(1.0, size=max_n), + "gue": lambda rep_rng, rep_i: gue_spacings(args.gue_matrix_size, max_n, rep_rng), + } + + for domain_name, builder in domain_builders.items(): + output["domains"][domain_name] = {} + output["summary"][domain_name] = {} + bases = [] + for rep_i in range(args.n_replicates): + rep_rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) + bases.append(builder(rep_rng, rep_i)) + + for n in sizes: + rows = [] + for rep_i, base in enumerate(bases): + rep_rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) + res = analyze_sequence( + base[:n], + alphas=alphas, + n_trials=args.n_trials, + n_baseline=args.n_baseline, + z_min=args.z_min, + rng=rep_rng, + ) + res["replicate"] = rep_i + rows.append(res) + summary = summarize_replicates(rows) + output["domains"][domain_name][str(n)] = rows + output["summary"][domain_name][str(n)] = summary + stable_rank = summary["stable_rank_mean"] + stable_text = f"{stable_rank:>11.3f}" if stable_rank is not None else f"{'NA':>11}" + print( + f"{domain_name:<22} {n:>6} {summary['rank_mean']:>7.3f} " + f"{summary['pc2_mean']:>7.3f} {summary['weak_observable_count_mean']:>5.2f} " + f"{stable_text}" + ) + + out_path = Path(args.out) + out_path.parent.mkdir(parents=True, exist_ok=True) + with out_path.open("w") as f: + json.dump(output, f, indent=2) + print(f"saved {out_path}") + return output + + +def main() -> None: + parser = argparse.ArgumentParser() + parser.add_argument("--sizes", default="128,256,512,1024,2048") + parser.add_argument("--n-replicates", type=int, default=8) + parser.add_argument("--gue-matrix-size", type=int, default=180) + parser.add_argument("--n-alpha", type=int, default=5) + parser.add_argument("--alpha-min", type=float, default=0.1) + parser.add_argument("--alpha-max", type=float, default=0.9) + parser.add_argument("--n-trials", type=int, default=8) + parser.add_argument("--n-baseline", type=int, default=16) + parser.add_argument("--z-min", type=float, default=2.0) + parser.add_argument("--seed", type=int, default=20260506) + parser.add_argument("--out", default="tools/data/perturbation_rank_size_curve.json") + args = parser.parse_args() + run(args) + + +if __name__ == "__main__": + main() codex Il risultato grezzo mostra già il punto: il rank “all-observables” resta alto anche in Poisson e shuffle-primi, ma quando si filtra sugli osservabili con segnale original-vs-shuffle sufficiente il rank scende vicino a 1. Ora estraggo le statistiche necessarie per il report e controllo i casi limite. exec /bin/bash -lc "python - <<'PY' import json from pathlib import Path p=Path('tools/data/perturbation_rank_size_curve.json') d=json.loads(p.read_text()) print('registry', d['observables_registry']) for domain, rows in d['summary'].items(): print('\\n'+domain) for n,s in rows.items(): print(n, 'rank', round(s['rank_mean'],3), '+/-', round(s['rank_std'],3), 'pc2', round(s['pc2_mean'],3), 'weak', round(s['weak_observable_count_mean'],2), 'stable_rank', None if s['stable_rank_mean'] is None else round(s['stable_rank_mean'],3), 'corr', round(s['rank_vs_weak_count_corr'],3)) PY" in /opt/MM_D-ND succeeded in 0ms: registry 1.0.0-2026-05-06 primes_windows 128 rank 1.789 +/- 0.469 pc2 0.155 weak 4.5 stable_rank 1.382 corr -0.085 256 rank 1.947 +/- 0.645 pc2 0.174 weak 4.75 stable_rank 1.262 corr -0.368 512 rank 1.892 +/- 0.372 pc2 0.142 weak 2.88 stable_rank 1.31 corr 0.337 1024 rank 1.679 +/- 0.409 pc2 0.117 weak 1.62 stable_rank 1.415 corr 0.453 2048 rank 1.442 +/- 0.213 pc2 0.081 weak 0.75 stable_rank 1.462 corr -0.375 prime_shuffle_control 128 rank 1.6 +/- 0.47 pc2 0.126 weak 4.75 stable_rank 1.174 corr -0.51 256 rank 1.952 +/- 0.411 pc2 0.183 weak 4.38 stable_rank 1.087 corr -0.049 512 rank 1.743 +/- 0.376 pc2 0.135 weak 4.12 stable_rank 1.216 corr -0.151 1024 rank 1.711 +/- 0.612 pc2 0.119 weak 4.38 stable_rank None corr -0.05 2048 rank 1.797 +/- 0.375 pc2 0.134 weak 3.62 stable_rank 1.428 corr -0.338 poisson 128 rank 1.765 +/- 0.689 pc2 0.133 weak 4.62 stable_rank 1.033 corr -0.191 256 rank 1.92 +/- 0.608 pc2 0.169 weak 4.5 stable_rank 1.177 corr 0.035 512 rank 1.737 +/- 0.676 pc2 0.126 weak 4.25 stable_rank 1.275 corr 0.208 1024 rank 1.881 +/- 0.517 pc2 0.176 weak 4.5 stable_rank 1.055 corr 0.512 2048 rank 1.952 +/- 0.499 pc2 0.175 weak 4.62 stable_rank 1.036 corr 0.315 gue 128 rank 1.703 +/- 0.348 pc2 0.126 weak 2.38 stable_rank 1.226 corr -0.089 256 rank 1.913 +/- 0.453 pc2 0.164 weak 2.25 stable_rank 1.141 corr 0.468 512 rank 1.542 +/- 0.313 pc2 0.111 weak 1.88 stable_rank 1.162 corr 0.087 1024 rank 1.551 +/- 0.395 pc2 0.105 weak 1.88 stable_rank 1.157 corr -0.229 2048 rank 1.234 +/- 0.224 pc2 0.046 weak 2.0 stable_rank 1.111 corr 0.0 exec /bin/bash -lc "python - <<'PY' import json, numpy as np from pathlib import Path d=json.loads(Path('tools/data/perturbation_rank_size_curve.json').read_text()) for domain in d['domains']: print('\\n'+domain) for n, reps in d['domains'][domain].items(): counts={obs:0 for obs in d['observables_used']} zvals={obs:[] for obs in d['observables_used']} for r in reps: for obs,z in r['original_vs_shuffle_z'].items(): zvals[obs].append(abs(z)) if abs(z)>=2: counts[obs]+=1 print(n, 'stable_counts', counts, 'mean_abs_z', {o:round(float(np.mean(v)),2) for o,v in zvals.items()}) PY" in /opt/MM_D-ND exec /bin/bash -lc "python - <<'PY' import json, numpy as np from pathlib import Path d=json.loads(Path('tools/data/perturbation_rank_size_curve.json').read_text()) xs=[]; ys=[] for domain, rows in d['domains'].items(): for n, reps in rows.items(): for r in reps: xs.append(r['weak_observable_count']) ys.append(r['pca_all_observables']['effective_rank']) print('global_rank_vs_weak_corr', float(np.corrcoef(xs,ys)[0,1])) for domain, rows in d['domains'].items(): xs=[]; ys=[] for reps in rows.values(): for r in reps: xs.append(r['weak_observable_count']); ys.append(r['pca_all_observables']['effective_rank']) print(domain, float(np.corrcoef(xs,ys)[0,1]) if np.std(xs)>1e-12 and np.std(ys)>1e-12 else 0.0) PY" in /opt/MM_D-ND diff --git a/tools/exp_perturbation_rank_size_curve.py b/tools/exp_perturbation_rank_size_curve.py new file mode 100644 index 0000000000000000000000000000000000000000..032847ca3d82ec3233ec2b41ad230dfe394d2791 --- /dev/null +++ b/tools/exp_perturbation_rank_size_curve.py @@ -0,0 +1,387 @@ +#!/usr/bin/env python3 +""" +exp_perturbation_rank_size_curve.py + +Reusable META audit for perturbation dimensionality. + +The 2026-05-06 06:25 cycle restricted the claim "GUE has a second +perturbation axis" to sample size, generator, and observable definitions. +This tool measures the size curve directly, using the canonical observable +registry and explicit original-vs-shuffle denominator diagnostics. + +The report owns interpretation. This script only measures: +- effective rank and PC2 across scale-selective perturbation profiles; +- original-vs-shuffle z-score per observable; +- whether apparent rank co-occurs with weak retention denominators. +""" + +from __future__ import annotations + +import argparse +import json +from pathlib import Path + +import numpy as np + +from observables_registry import ( + OBSERVABLES_CANONICAL, + OBSERVABLES_REGISTRY_VERSION, + compute_canonical, +) + + +OBS_NAMES = list(OBSERVABLES_CANONICAL.keys()) +PERT_NAMES = ["adjacent_swap", "block_shuffle", "large_gap_only", "uniform"] + + +def prime_gaps(n_gaps: int) -> np.ndarray: + """Return the first n_gaps prime gaps using a compact numpy sieve.""" + limit = max(100, int(n_gaps * (np.log(n_gaps + 10) + np.log(np.log(n_gaps + 10)) + 5))) + while True: + sieve = np.ones(limit + 1, dtype=bool) + sieve[:2] = False + for p in range(2, int(limit**0.5) + 1): + if sieve[p]: + sieve[p * p : limit + 1 : p] = False + primes = np.flatnonzero(sieve) + if len(primes) >= n_gaps + 1: + return np.diff(primes[: n_gaps + 1]).astype(float) + limit *= 2 + + +def gue_spacings(matrix_size: int, min_spacings: int, rng: np.random.Generator) -> np.ndarray: + """Generate unfolded GUE spacings by concatenating independent matrices.""" + parts = [] + edge = max(2, matrix_size // 10) + while sum(len(x) for x in parts) < min_spacings: + real = rng.standard_normal((matrix_size, matrix_size)) + imag = rng.standard_normal((matrix_size, matrix_size)) + h = real + 1j * imag + h = (h + h.conj().T) / (2.0 * np.sqrt(matrix_size)) + eigs = np.sort(np.linalg.eigvalsh(h).real) + bulk = eigs[edge:-edge] + gaps = np.diff(bulk) + mean = float(np.mean(gaps)) + if mean > 1e-15: + parts.append(gaps / mean) + return np.concatenate(parts)[:min_spacings].astype(float) + + +def perturb_adjacent_swap(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + idx = np.arange(0, len(out) - 1, 2) + chosen = idx[rng.random(len(idx)) < alpha] + tmp = out[chosen].copy() + out[chosen] = out[chosen + 1] + out[chosen + 1] = tmp + return out + + +def perturb_block_shuffle(gaps: np.ndarray, alpha: float, rng: np.random.Generator, block_size: int = 64) -> np.ndarray: + out = gaps.copy() + n_blocks = len(out) // block_size + if n_blocks <= 0: + return rng.permutation(out) if alpha > 0 else out + k = int(round(alpha * n_blocks)) + if k <= 0: + return out + for block in rng.choice(n_blocks, size=min(k, n_blocks), replace=False): + start = block * block_size + end = min(start + block_size, len(out)) + rng.shuffle(out[start:end]) + return out + + +def perturb_large_gap_only(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + idx = np.flatnonzero(out > np.median(out)) + k = int(round(alpha * len(idx))) + if k < 2: + return out + chosen = rng.choice(idx, size=min(k, len(idx)), replace=False) + vals = out[chosen].copy() + rng.shuffle(vals) + out[chosen] = vals + return out + + +def perturb_uniform(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + k = int(round(alpha * len(out))) + if k < 2: + return out + chosen = rng.choice(len(out), size=min(k, len(out)), replace=False) + vals = out[chosen].copy() + rng.shuffle(vals) + out[chosen] = vals + return out + + +PERTURB = { + "adjacent_swap": perturb_adjacent_swap, + "block_shuffle": perturb_block_shuffle, + "large_gap_only": perturb_large_gap_only, + "uniform": perturb_uniform, +} + + +def pca_summary(vectors: list[list[float]], labels: list[str]) -> dict: + matrix = np.array(vectors, dtype=float) + matrix = matrix - np.mean(matrix, axis=0, keepdims=True) + if min(matrix.shape) == 0 or np.max(np.abs(matrix)) <= 1e-15: + return { + "explained_variance": [], + "effective_rank": 0.0, + "centroid_cosine": {}, + "pc2": 0.0, + } + _, singular, _ = np.linalg.svd(matrix, full_matrices=False) + energy = singular * singular + if float(np.sum(energy)) <= 1e-15: + explained = np.zeros_like(energy) + effective_rank = 0.0 + else: + explained = energy / np.sum(energy) + pos = explained[explained > 1e-15] + effective_rank = float(np.exp(-np.sum(pos * np.log(pos)))) + + centroids = {} + for name in PERT_NAMES: + vals = np.array([v for v, label in zip(vectors, labels) if label == name], dtype=float) + if len(vals): + centroids[name] = np.mean(vals, axis=0) + + cosine = {} + for i, a_name in enumerate(PERT_NAMES): + for b_name in PERT_NAMES[i + 1 :]: + if a_name not in centroids or b_name not in centroids: + continue + a = centroids[a_name] + b = centroids[b_name] + denom = np.linalg.norm(a) * np.linalg.norm(b) + cosine[f"{a_name}_vs_{b_name}"] = float(np.dot(a, b) / denom) if denom > 1e-15 else 0.0 + + return { + "explained_variance": [float(x) for x in explained], + "effective_rank": effective_rank, + "centroid_cosine": cosine, + "pc2": float(explained[1]) if len(explained) > 1 else 0.0, + } + + +def analyze_sequence( + gaps: np.ndarray, + alphas: list[float], + n_trials: int, + n_baseline: int, + z_min: float, + rng: np.random.Generator, +) -> dict: + original = compute_canonical(gaps) + baseline_vals = {obs: [] for obs in OBS_NAMES} + for _ in range(n_baseline): + obs = compute_canonical(rng.permutation(gaps)) + for name in OBS_NAMES: + baseline_vals[name].append(obs[name]) + + baseline = {} + z = {} + denom = {} + for name in OBS_NAMES: + vals = np.array(baseline_vals[name], dtype=float) + mean = float(np.mean(vals)) + std = float(np.std(vals, ddof=1)) if len(vals) > 1 else 0.0 + baseline[name] = {"mean": mean, "std": std} + denom[name] = float(original[name] - mean) + z[name] = float(denom[name] / std) if std > 1e-15 else 0.0 + + stable_obs = [name for name in OBS_NAMES if abs(z[name]) >= z_min] + all_vectors = [] + screened_vectors = [] + labels = [] + profiles = [] + + for pert_name in PERT_NAMES: + for alpha in alphas: + trial_vals = {obs: [] for obs in OBS_NAMES} + for _ in range(n_trials): + perturbed = PERTURB[pert_name](gaps, alpha, rng) + obs = compute_canonical(perturbed) + for name in OBS_NAMES: + trial_vals[name].append(obs[name]) + means = {name: float(np.mean(trial_vals[name])) for name in OBS_NAMES} + retention = {} + for name in OBS_NAMES: + retention[name] = ( + float((means[name] - baseline[name]["mean"]) / denom[name]) + if abs(denom[name]) > 1e-12 + else 0.0 + ) + all_vector = [retention[name] for name in OBS_NAMES] + screened_vector = [retention[name] for name in stable_obs] + all_vectors.append(all_vector) + if len(stable_obs) >= 2: + screened_vectors.append(screened_vector) + labels.append(pert_name) + profiles.append( + { + "perturbation": pert_name, + "alpha": float(alpha), + "retention": retention, + "retention_vector": all_vector, + } + ) + + all_pca = pca_summary(all_vectors, labels) + screened_pca = pca_summary(screened_vectors, labels) if len(stable_obs) >= 2 else None + + return { + "n_gaps": int(len(gaps)), + "original": original, + "full_shuffle_baseline": baseline, + "denominator": denom, + "original_vs_shuffle_z": z, + "stable_observables": stable_obs, + "weak_observable_count": int(len(OBS_NAMES) - len(stable_obs)), + "profiles": profiles, + "pca_all_observables": all_pca, + "pca_stable_observables": screened_pca, + } + + +def summarize_replicates(items: list[dict]) -> dict: + def arr(path: tuple[str, ...]) -> np.ndarray: + vals = [] + for item in items: + x = item + for key in path: + if x is None: + break + x = x.get(key) + if isinstance(x, (int, float)): + vals.append(float(x)) + return np.array(vals, dtype=float) + + rank = arr(("pca_all_observables", "effective_rank")) + pc2 = arr(("pca_all_observables", "pc2")) + weak = np.array([item["weak_observable_count"] for item in items], dtype=float) + stable_rank = arr(("pca_stable_observables", "effective_rank")) + cos = arr(("pca_all_observables", "centroid_cosine", "adjacent_swap_vs_large_gap_only")) + + out = { + "n_replicates": len(items), + "rank_mean": float(np.mean(rank)) if len(rank) else 0.0, + "rank_std": float(np.std(rank, ddof=1)) if len(rank) > 1 else 0.0, + "pc2_mean": float(np.mean(pc2)) if len(pc2) else 0.0, + "pc2_std": float(np.std(pc2, ddof=1)) if len(pc2) > 1 else 0.0, + "weak_observable_count_mean": float(np.mean(weak)) if len(weak) else 0.0, + "stable_rank_mean": float(np.mean(stable_rank)) if len(stable_rank) else None, + "stable_rank_std": float(np.std(stable_rank, ddof=1)) if len(stable_rank) > 1 else 0.0, + "adjacent_vs_large_cosine_mean": float(np.mean(cos)) if len(cos) else 0.0, + "adjacent_vs_large_cosine_std": float(np.std(cos, ddof=1)) if len(cos) > 1 else 0.0, + } + if len(rank) > 1 and np.std(weak) > 1e-15 and np.std(rank) > 1e-15: + out["rank_vs_weak_count_corr"] = float(np.corrcoef(rank, weak)[0, 1]) + else: + out["rank_vs_weak_count_corr"] = 0.0 + return out + + +def build_prime_windows(max_n: int, n_reps: int) -> list[np.ndarray]: + total = max_n * n_reps + max_n + gaps = prime_gaps(total) + max_start = len(gaps) - max_n + starts = np.linspace(0, max_start, n_reps, dtype=int) + return [gaps[start : start + max_n].astype(float) for start in starts] + + +def run(args: argparse.Namespace) -> dict: + root_rng = np.random.default_rng(args.seed) + sizes = [int(x) for x in args.sizes.split(",") if x.strip()] + max_n = max(sizes) + alphas = [float(x) for x in np.linspace(args.alpha_min, args.alpha_max, args.n_alpha)] + + output = { + "experiment": "perturbation_rank_size_curve", + "question": "Does perturbation effective-rank stabilize with sample size under canonical observables?", + "observables_registry": OBSERVABLES_REGISTRY_VERSION, + "observables_used": OBS_NAMES, + "perturbations": PERT_NAMES, + "params": vars(args), + "alphas": alphas, + "domains": {}, + "summary": {}, + } + + prime_windows = build_prime_windows(max_n, args.n_replicates) + + print(f"observables_registry={OBSERVABLES_REGISTRY_VERSION}") + print(f"{'domain':<22} {'N':>6} {'rank':>7} {'PC2':>7} {'weak':>5} {'stable_rank':>11}") + + domain_builders = { + "primes_windows": lambda rep_rng, rep_i: prime_windows[rep_i], + "prime_shuffle_control": lambda rep_rng, rep_i: rep_rng.permutation(prime_windows[rep_i]), + "poisson": lambda rep_rng, rep_i: rep_rng.exponential(1.0, size=max_n), + "gue": lambda rep_rng, rep_i: gue_spacings(args.gue_matrix_size, max_n, rep_rng), + } + + for domain_name, builder in domain_builders.items(): + output["domains"][domain_name] = {} + output["summary"][domain_name] = {} + bases = [] + for rep_i in range(args.n_replicates): + rep_rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) + bases.append(builder(rep_rng, rep_i)) + + for n in sizes: + rows = [] + for rep_i, base in enumerate(bases): + rep_rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) + res = analyze_sequence( + base[:n], + alphas=alphas, + n_trials=args.n_trials, + n_baseline=args.n_baseline, + z_min=args.z_min, + rng=rep_rng, + ) + res["replicate"] = rep_i + rows.append(res) + summary = summarize_replicates(rows) + output["domains"][domain_name][str(n)] = rows + output["summary"][domain_name][str(n)] = summary + stable_rank = summary["stable_rank_mean"] + stable_text = f"{stable_rank:>11.3f}" if stable_rank is not None else f"{'NA':>11}" + print( + f"{domain_name:<22} {n:>6} {summary['rank_mean']:>7.3f} " + f"{summary['pc2_mean']:>7.3f} {summary['weak_observable_count_mean']:>5.2f} " + f"{stable_text}" + ) + + out_path = Path(args.out) + out_path.parent.mkdir(parents=True, exist_ok=True) + with out_path.open("w") as f: + json.dump(output, f, indent=2) + print(f"saved {out_path}") + return output + + +def main() -> None: + parser = argparse.ArgumentParser() + parser.add_argument("--sizes", default="128,256,512,1024,2048") + parser.add_argument("--n-replicates", type=int, default=8) + parser.add_argument("--gue-matrix-size", type=int, default=180) + parser.add_argument("--n-alpha", type=int, default=5) + parser.add_argument("--alpha-min", type=float, default=0.1) + parser.add_argument("--alpha-max", type=float, default=0.9) + parser.add_argument("--n-trials", type=int, default=8) + parser.add_argument("--n-baseline", type=int, default=16) + parser.add_argument("--z-min", type=float, default=2.0) + parser.add_argument("--seed", type=int, default=20260506) + parser.add_argument("--out", default="tools/data/perturbation_rank_size_curve.json") + args = parser.parse_args() + run(args) + + +if __name__ == "__main__": + main() succeeded in 151ms: global_rank_vs_weak_corr 0.2133273037790684 primes_windows 0.3104375529680552 prime_shuffle_control -0.21600043082167658 poisson 0.15479027162735615 gue 0.2589384560147655 succeeded in 156ms: primes_windows 128 stable_counts {'SR': 2, 'SR2': 0, 'L1': 0, 'L2': 1, 'triple_var': 1} mean_abs_z {'SR': 1.57, 'SR2': 0.93, 'L1': 0.68, 'L2': 0.61, 'triple_var': 0.67} 256 stable_counts {'SR': 1, 'SR2': 0, 'L1': 1, 'L2': 0, 'triple_var': 0} mean_abs_z {'SR': 1.77, 'SR2': 0.82, 'L1': 1.08, 'L2': 0.73, 'triple_var': 1.13} 512 stable_counts {'SR': 8, 'SR2': 1, 'L1': 4, 'L2': 1, 'triple_var': 3} mean_abs_z {'SR': 3.15, 'SR2': 1.39, 'L1': 2.18, 'L2': 1.43, 'triple_var': 2.17} 1024 stable_counts {'SR': 8, 'SR2': 4, 'L1': 5, 'L2': 3, 'triple_var': 7} mean_abs_z {'SR': 3.58, 'SR2': 1.63, 'L1': 2.21, 'L2': 1.62, 'triple_var': 2.71} 2048 stable_counts {'SR': 8, 'SR2': 7, 'L1': 8, 'L2': 3, 'triple_var': 8} mean_abs_z {'SR': 5.19, 'SR2': 2.63, 'L1': 3.96, 'L2': 1.78, 'triple_var': 4.37} prime_shuffle_control 128 stable_counts {'SR': 0, 'SR2': 1, 'L1': 0, 'L2': 1, 'triple_var': 0} mean_abs_z {'SR': 0.79, 'SR2': 1.38, 'L1': 0.6, 'L2': 1.45, 'triple_var': 0.56} 256 stable_counts {'SR': 1, 'SR2': 3, 'L1': 0, 'L2': 1, 'triple_var': 0} mean_abs_z {'SR': 1.22, 'SR2': 1.77, 'L1': 0.7, 'L2': 1.21, 'triple_var': 0.73} 512 stable_counts {'SR': 2, 'SR2': 1, 'L1': 1, 'L2': 1, 'triple_var': 2} mean_abs_z {'SR': 0.97, 'SR2': 1.1, 'L1': 0.98, 'L2': 1.13, 'triple_var': 1.03} 1024 stable_counts {'SR': 1, 'SR2': 0, 'L1': 0, 'L2': 3, 'triple_var': 1} mean_abs_z {'SR': 1.11, 'SR2': 0.85, 'L1': 0.94, 'L2': 1.39, 'triple_var': 1.08} 2048 stable_counts {'SR': 3, 'SR2': 1, 'L1': 2, 'L2': 2, 'triple_var': 3} mean_abs_z {'SR': 1.65, 'SR2': 0.91, 'L1': 1.48, 'L2': 1.28, 'triple_var': 1.69} poisson 128 stable_counts {'SR': 0, 'SR2': 0, 'L1': 2, 'L2': 0, 'triple_var': 1} mean_abs_z {'SR': 0.91, 'SR2': 0.66, 'L1': 1.13, 'L2': 0.53, 'triple_var': 1.34} 256 stable_counts {'SR': 1, 'SR2': 1, 'L1': 1, 'L2': 0, 'triple_var': 1} mean_abs_z {'SR': 0.68, 'SR2': 1.06, 'L1': 0.83, 'L2': 0.78, 'triple_var': 0.84} 512 stable_counts {'SR': 1, 'SR2': 2, 'L1': 1, 'L2': 0, 'triple_var': 2} mean_abs_z {'SR': 0.84, 'SR2': 1.08, 'L1': 0.78, 'L2': 1.04, 'triple_var': 0.91} 1024 stable_counts {'SR': 0, 'SR2': 2, 'L1': 1, 'L2': 0, 'triple_var': 1} mean_abs_z {'SR': 0.76, 'SR2': 1.19, 'L1': 0.63, 'L2': 0.58, 'triple_var': 0.79} 2048 stable_counts {'SR': 0, 'SR2': 1, 'L1': 1, 'L2': 0, 'triple_var': 1} mean_abs_z {'SR': 0.62, 'SR2': 1.12, 'L1': 0.91, 'L2': 0.59, 'triple_var': 0.93} gue 128 stable_counts {'SR': 4, 'SR2': 0, 'L1': 8, 'L2': 1, 'triple_var': 8} mean_abs_z {'SR': 2.02, 'SR2': 0.71, 'L1': 3.72, 'L2': 0.73, 'triple_var': 3.38} 256 stable_counts {'SR': 4, 'SR2': 1, 'L1': 8, 'L2': 1, 'triple_var': 8} mean_abs_z {'SR': 2.9, 'SR2': 0.96, 'L1': 4.43, 'L2': 1.18, 'triple_var': 4.03} 512 stable_counts {'SR': 8, 'SR2': 1, 'L1': 8, 'L2': 0, 'triple_var': 8} mean_abs_z {'SR': 4.11, 'SR2': 0.91, 'L1': 5.76, 'L2': 0.99, 'triple_var': 5.35} 1024 stable_counts {'SR': 8, 'SR2': 1, 'L1': 8, 'L2': 0, 'triple_var': 8} mean_abs_z {'SR': 5.78, 'SR2': 0.98, 'L1': 8.36, 'L2': 0.93, 'triple_var': 7.48} 2048 stable_counts {'SR': 8, 'SR2': 0, 'L1': 8, 'L2': 0, 'triple_var': 8} mean_abs_z {'SR': 8.38, 'SR2': 0.67, 'L1': 11.58, 'L2': 0.89, 'triple_var': 11.66} exec /bin/bash -lc 'git status --short' in /opt/MM_D-ND succeeded in 0ms: M applications/INDEX.draft.json M applications/soluzioni/20260430_0330_three-regimes-of-modular-memory-algebraic/finding_index.draft.json M applications/soluzioni/20260430_0330_three-regimes-of-modular-memory-algebraic/manifest.draft.json M tools/LAB_AGENT_CONTEXT.md M tools/data/agent_field_live.md M tools/data/conoscenza_teorie.json M tools/data/consecutio.json M tools/data/consecutio_processata.json M tools/data/domande_fondamentali.json M tools/data/incrocio_risultato.json M tools/data/lab_data.json M tools/data/lab_graph.json M tools/data/lab_health.json M tools/data/lab_session_log.jsonl M tools/data/pipeline_state.json M tools/data/ponti_evoluti.json M tools/data/refresh_detector_state.json M tools/data/seme.json D tools/data/seme_archive/piano_37.json D tools/data/seme_archive/piano_38.json D tools/data/seme_archive/piano_4.json D tools/data/seme_archive/piano_5.json D tools/data/seme_archive/piano_6.json M tools/data/seme_axioms.json M tools/data/seme_backup_pre_run.json M tools/data/tm1_figures/tensions.json M tools/data/tm1_figures/tensions_raw.json M tools/lab_agent.sh M tools/lab_autopsy.py ?? applications/published/20260504_0901_the-two-markov-layers-are-coupled/ ?? applications/published/20260504_1219_markov-layers-pass-the-first-recovery/ ?? applications/published/20260505_0330_observable-rank-audit-many-probes-one/ ?? applications/scoperte/20260502_0330_the-dipolar-phase-transition-is-nearly_auto/ ?? applications/scoperte/20260503_0330_markov-memory-has-two-visible-layers_auto/ ?? applications/scoperte/20260504_0901_the-two-markov-layers-are-coupled_auto/ ?? applications/scoperte/20260504_1219_markov-layers-pass-the-first-recovery_auto/ ?? applications/scoperte/20260505_0330_observable-rank-audit-many-probes-one_auto/ ?? applications/scoperte/20260506_0625_perturbation-dimensionality-is-not-yet-a_auto/ ?? applications/soluzioni/20260502_0330_the-dipolar-phase-transition-is-nearly/ ?? applications/soluzioni/20260503_0330_markov-memory-has-two-visible-layers/ ?? applications/soluzioni/20260504_0901_the-two-markov-layers-are-coupled/ ?? applications/soluzioni/20260504_1219_markov-layers-pass-the-first-recovery/ ?? applications/soluzioni/20260505_0330_observable-rank-audit-many-probes-one/ ?? findings_gallery.md ?? tools/data/.last_telegram_msg ?? tools/data/3d_boundary_layers.json ?? tools/data/aeternitas/ ?? tools/data/alignment_active.json ?? tools/data/alignment_markers.jsonl ?? tools/data/bicono_projections.jsonl ?? tools/data/boundary_coherence.json ?? tools/data/boundary_shuffle_audit.json ?? tools/data/brody_calibration_results.json ?? tools/data/brody_flow.json ?? tools/data/conoscenza_teorie.json.bak.retraction_22_04 ?? tools/data/cross_domain_dipolar_direction.json ?? tools/data/cross_observable_consistency.json ?? tools/data/crossover_phase_test.json ?? tools/data/dipolar_crossover.json ?? tools/data/dipolar_vector_scaling.json ?? tools/data/domandatore/domandatore_20260421_0746.json ?? tools/data/domandatore/domandatore_20260422_0345.json ?? tools/data/domandatore/domandatore_20260423_0345.json ?? tools/data/domandatore/domandatore_20260424_0345.json ?? tools/data/domandatore/domandatore_20260425_0345.json ?? tools/data/domandatore/domandatore_20260426_0345.json ?? tools/data/domandatore/domandatore_20260427_0345.json ?? tools/data/domandatore/domandatore_20260428_0345.json ?? tools/data/domandatore/domandatore_20260428_1236.json ?? tools/data/domandatore/domandatore_20260429_0345.json ?? tools/data/domandatore/domandatore_20260430_0345.json ?? tools/data/domandatore/domandatore_20260501_0345.json ?? tools/data/domandatore/domandatore_20260502_0345.json ?? tools/data/domandatore/domandatore_20260503_0345.json ?? tools/data/domandatore/domandatore_20260504_0345.json ?? tools/data/domandatore/domandatore_20260505_0345.json ?? tools/data/domandatore/domandatore_20260506_0345.json ?? tools/data/evolution/evolution_20260422_0330.md ?? tools/data/evolution/evolution_20260422_1616.md ?? tools/data/evolution/evolution_20260423_0330.md ?? tools/data/evolution/evolution_20260424_0330.md ?? tools/data/evolution/evolution_20260425_0330.md ?? tools/data/evolution/evolution_20260427_0330.md ?? tools/data/evolution/evolution_20260428_0330.md ?? tools/data/evolution/evolution_20260429_0852.md ?? tools/data/evolution/evolution_20260429_1013.md ?? tools/data/evolution/evolution_20260430_0330.md ?? tools/data/evolution/evolution_20260430_1905.md ?? tools/data/evolution/evolution_20260430_1946.md ?? tools/data/evolution/evolution_20260501_0330.md ?? tools/data/evolution/evolution_20260501_0725.md ?? tools/data/evolution/evolution_20260501_0931.md ?? tools/data/evolution/evolution_20260502_0330.md ?? tools/data/evolution/evolution_20260503_0330.md ?? tools/data/evolution/evolution_20260504_0901.md ?? tools/data/evolution/evolution_20260504_1219.md ?? tools/data/evolution/evolution_20260505_0330.md ?? tools/data/evolution/evolution_20260506_0330.md ?? tools/data/incrocio_20260422_0336.json ?? tools/data/incrocio_20260423_0335.json ?? tools/data/incrocio_20260424_0347.json ?? tools/data/incrocio_20260425_0339.json ?? tools/data/incrocio_20260428_0340.json ?? tools/data/incrocio_20260429_0859.json ?? tools/data/incrocio_20260506_0633.json ?? tools/data/lab_bridge_issues.jsonl ?? tools/data/magnitude_psd_from_acf.json ?? tools/data/markov3_observable_hunt.json ?? tools/data/markov_dipolar_decomposition.json ?? tools/data/markov_k_direction.json ?? tools/data/markov_layer_recovery_audit.json ?? tools/data/markov_memory_by_gue_type.json ?? tools/data/markov_scale_function.json ?? tools/data/meta_tautology_test.json ?? tools/data/mod3_scaling.json ?? tools/data/mod3_vs_residual_ordering.json ?? tools/data/modular_algebra_depth.json ?? tools/data/modular_memory_spectrum.json ?? tools/data/observable_rank_audit.json ?? tools/data/observable_rank_audit_seed20260506.json ?? tools/data/perturbation_dimensionality_audit.json ?? tools/data/perturbation_dimensionality_audit_scale0330.json ?? tools/data/perturbation_rank_size_curve.json ?? tools/data/promotions/ ?? tools/data/reports/_quarantine_falsifier_29_04/ ?? tools/data/reports/agent_20260422_0330.md ?? tools/data/reports/agent_20260422_1616.md ?? tools/data/reports/agent_20260423_0330.md ?? tools/data/reports/agent_20260424_0330.md ?? tools/data/reports/agent_20260425_0330.md ?? tools/data/reports/agent_20260426_0330.md ?? tools/data/reports/agent_20260427_0330.md ?? tools/data/reports/agent_20260428_0330.md ?? tools/data/reports/agent_20260429_1013.md ?? tools/data/reports/agent_20260429_1041.md ?? tools/data/reports/agent_20260430_0330.md ?? tools/data/reports/agent_20260430_1905.md ?? tools/data/reports/agent_20260430_1919.md ?? tools/data/reports/agent_20260430_1946.md ?? tools/data/reports/agent_20260501_0330.md ?? tools/data/reports/agent_20260501_0725.md ?? tools/data/reports/agent_20260501_0931.md ?? tools/data/reports/agent_20260502_0330.md ?? tools/data/reports/agent_20260503_0330.md ?? tools/data/reports/agent_20260504_0901.md ?? tools/data/reports/agent_20260504_1219.md ?? tools/data/reports/agent_20260505_0330.md ?? tools/data/reports/agent_20260505_1022.md ?? tools/data/reports/agent_20260506_0330.md ?? tools/data/reports/agent_20260506_0625.md ?? tools/data/reports/ddf_20260422_0405.json ?? tools/data/reports/ddf_20260423_0405.json ?? tools/data/reports/ddf_20260424_0405.json ?? tools/data/reports/ddf_20260425_0405.json ?? tools/data/reports/ddf_20260426_0405.json ?? tools/data/reports/ddf_20260427_0405.json ?? tools/data/reports/ddf_20260428_0405.json ?? tools/data/reports/ddf_20260429_0405.json ?? tools/data/reports/ddf_20260430_0405.json ?? tools/data/reports/ddf_20260501_0405.json ?? tools/data/reports/ddf_20260502_0405.json ?? tools/data/reports/ddf_20260503_0405.json ?? tools/data/reports/ddf_20260504_0405.json ?? tools/data/reports/ddf_20260505_0405.json ?? tools/data/reports/ddf_20260505_0636.json ?? tools/data/reports/ddf_20260506_0405.json ?? tools/data/reports/evolution_20260422_0330.md ?? tools/data/reports/evolution_20260422_1616.md ?? tools/data/reports/evolution_20260423_0330.md ?? tools/data/reports/evolution_20260424_0330.md ?? tools/data/reports/evolution_20260425_0330.md ?? tools/data/reports/evolution_20260427_0330.md ?? tools/data/reports/evolution_20260428_0330.md ?? tools/data/reports/evolution_20260503_0330.md ?? tools/data/reports/evolution_20260504_0330.md ?? tools/data/reports/evolution_20260505_0330.md ?? tools/data/reports/evolution_20260506_0330.md ?? tools/data/reports/falsifier_20260429_1013.json ?? tools/data/reports/falsifier_20260429_1041.json ?? tools/data/reports/falsifier_20260430_0330.json ?? tools/data/reports/falsifier_20260430_1905.json ?? tools/data/reports/falsifier_20260430_1919.json ?? tools/data/reports/falsifier_20260430_1946.json ?? tools/data/reports/falsifier_20260501_0330.json ?? tools/data/reports/falsifier_20260501_0725.json ?? tools/data/reports/falsifier_20260501_0931.json ?? tools/data/reports/falsifier_20260502_0330.json ?? tools/data/reports/falsifier_20260503_0330.json ?? tools/data/reports/falsifier_20260504_0901.json ?? tools/data/reports/falsifier_20260504_1219.json ?? tools/data/reports/falsifier_20260505_0330.json ?? tools/data/reports/falsifier_20260506_0330.raw.txt ?? tools/data/reports/falsifier_20260506_0625.json ?? tools/data/reports/incident_20260504_0721.md ?? tools/data/reports/incident_20260504_1138.md ?? tools/data/reports/latest.md ?? tools/data/reports/trajectory_apply_20260506_1931.json ?? tools/data/reports/trajectory_apply_20260506_1941.json ?? tools/data/scale_selective_perturbation.json ?? tools/data/selective_layer_decoupling.json ?? tools/data/seme_archive/piano_61.json ?? tools/data/seme_archive/piano_62.json ?? tools/data/seme_archive/piano_63.json ?? tools/data/seme_archive/piano_64.json ?? tools/data/seme_archive/piano_65.json ?? tools/data/spectral_rigidity_results.json ?? tools/data/two_channel_boundary.json ?? tools/data/two_channel_cross_domain.json ?? tools/data/two_channel_shuffle_audit.json ?? tools/data/two_layer_universality.json ?? tools/data/valutatore_log.jsonl ?? tools/data/veritas/ ?? tools/exp_3d_boundary_layers.py ?? tools/exp_boundary_coherence.py ?? tools/exp_boundary_shuffle_audit.py ?? tools/exp_brody_calibration.py ?? tools/exp_brody_flow.py ?? tools/exp_cross_domain_dipolar_direction.py ?? tools/exp_cross_observable_consistency.py ?? tools/exp_crossover_phase_test.py ?? tools/exp_dipolar_angle_reference.py ?? tools/exp_dipolar_crossover.py ?? tools/exp_dipolar_vector_scaling.py ?? tools/exp_magnitude_psd_from_acf.py ?? tools/exp_markov3_observable_hunt.py ?? tools/exp_markov_dipolar_decomposition.py ?? tools/exp_markov_k_direction.py ?? tools/exp_markov_layer_recovery_audit.py ?? tools/exp_markov_memory_by_gue_type.py ?? tools/exp_markov_psd_prediction.py ?? tools/exp_markov_scale_function.py ?? tools/exp_meta_tautology_test.py ?? tools/exp_mod3_scaling.py ?? tools/exp_mod3_vs_residual_ordering.py ?? tools/exp_modular_algebra_depth.py ?? tools/exp_modular_memory_spectrum.py ?? tools/exp_observable_rank_audit.py ?? tools/exp_perturbation_dimensionality_audit.py ?? tools/exp_perturbation_rank_size_curve.py ?? tools/exp_scale_selective_perturbation.py ?? tools/exp_selective_layer_decoupling.py ?? tools/exp_spectral_rigidity.py ?? tools/exp_two_channel_boundary.py ?? tools/exp_two_channel_cross_domain.py ?? tools/exp_two_channel_shuffle_audit.py ?? tools/exp_two_layer_universality.py ?? tools/lab_trajectory_apply.py ?? tools/observables_registry.py exec /bin/bash -lc 'python -m py_compile tools/exp_perturbation_rank_size_curve.py tools/observables_registry.py' in /opt/MM_D-ND succeeded in 0ms: diff --git a/tools/exp_perturbation_rank_size_curve.py b/tools/exp_perturbation_rank_size_curve.py new file mode 100644 index 0000000000000000000000000000000000000000..032847ca3d82ec3233ec2b41ad230dfe394d2791 --- /dev/null +++ b/tools/exp_perturbation_rank_size_curve.py @@ -0,0 +1,387 @@ +#!/usr/bin/env python3 +""" +exp_perturbation_rank_size_curve.py + +Reusable META audit for perturbation dimensionality. + +The 2026-05-06 06:25 cycle restricted the claim "GUE has a second +perturbation axis" to sample size, generator, and observable definitions. +This tool measures the size curve directly, using the canonical observable +registry and explicit original-vs-shuffle denominator diagnostics. + +The report owns interpretation. This script only measures: +- effective rank and PC2 across scale-selective perturbation profiles; +- original-vs-shuffle z-score per observable; +- whether apparent rank co-occurs with weak retention denominators. +""" + +from __future__ import annotations + +import argparse +import json +from pathlib import Path + +import numpy as np + +from observables_registry import ( + OBSERVABLES_CANONICAL, + OBSERVABLES_REGISTRY_VERSION, + compute_canonical, +) + + +OBS_NAMES = list(OBSERVABLES_CANONICAL.keys()) +PERT_NAMES = ["adjacent_swap", "block_shuffle", "large_gap_only", "uniform"] + + +def prime_gaps(n_gaps: int) -> np.ndarray: + """Return the first n_gaps prime gaps using a compact numpy sieve.""" + limit = max(100, int(n_gaps * (np.log(n_gaps + 10) + np.log(np.log(n_gaps + 10)) + 5))) + while True: + sieve = np.ones(limit + 1, dtype=bool) + sieve[:2] = False + for p in range(2, int(limit**0.5) + 1): + if sieve[p]: + sieve[p * p : limit + 1 : p] = False + primes = np.flatnonzero(sieve) + if len(primes) >= n_gaps + 1: + return np.diff(primes[: n_gaps + 1]).astype(float) + limit *= 2 + + +def gue_spacings(matrix_size: int, min_spacings: int, rng: np.random.Generator) -> np.ndarray: + """Generate unfolded GUE spacings by concatenating independent matrices.""" + parts = [] + edge = max(2, matrix_size // 10) + while sum(len(x) for x in parts) < min_spacings: + real = rng.standard_normal((matrix_size, matrix_size)) + imag = rng.standard_normal((matrix_size, matrix_size)) + h = real + 1j * imag + h = (h + h.conj().T) / (2.0 * np.sqrt(matrix_size)) + eigs = np.sort(np.linalg.eigvalsh(h).real) + bulk = eigs[edge:-edge] + gaps = np.diff(bulk) + mean = float(np.mean(gaps)) + if mean > 1e-15: + parts.append(gaps / mean) + return np.concatenate(parts)[:min_spacings].astype(float) + + +def perturb_adjacent_swap(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + idx = np.arange(0, len(out) - 1, 2) + chosen = idx[rng.random(len(idx)) < alpha] + tmp = out[chosen].copy() + out[chosen] = out[chosen + 1] + out[chosen + 1] = tmp + return out + + +def perturb_block_shuffle(gaps: np.ndarray, alpha: float, rng: np.random.Generator, block_size: int = 64) -> np.ndarray: + out = gaps.copy() + n_blocks = len(out) // block_size + if n_blocks <= 0: + return rng.permutation(out) if alpha > 0 else out + k = int(round(alpha * n_blocks)) + if k <= 0: + return out + for block in rng.choice(n_blocks, size=min(k, n_blocks), replace=False): + start = block * block_size + end = min(start + block_size, len(out)) + rng.shuffle(out[start:end]) + return out + + +def perturb_large_gap_only(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + idx = np.flatnonzero(out > np.median(out)) + k = int(round(alpha * len(idx))) + if k < 2: + return out + chosen = rng.choice(idx, size=min(k, len(idx)), replace=False) + vals = out[chosen].copy() + rng.shuffle(vals) + out[chosen] = vals + return out + + +def perturb_uniform(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + k = int(round(alpha * len(out))) + if k < 2: + return out + chosen = rng.choice(len(out), size=min(k, len(out)), replace=False) + vals = out[chosen].copy() + rng.shuffle(vals) + out[chosen] = vals + return out + + +PERTURB = { + "adjacent_swap": perturb_adjacent_swap, + "block_shuffle": perturb_block_shuffle, + "large_gap_only": perturb_large_gap_only, + "uniform": perturb_uniform, +} + + +def pca_summary(vectors: list[list[float]], labels: list[str]) -> dict: + matrix = np.array(vectors, dtype=float) + matrix = matrix - np.mean(matrix, axis=0, keepdims=True) + if min(matrix.shape) == 0 or np.max(np.abs(matrix)) <= 1e-15: + return { + "explained_variance": [], + "effective_rank": 0.0, + "centroid_cosine": {}, + "pc2": 0.0, + } + _, singular, _ = np.linalg.svd(matrix, full_matrices=False) + energy = singular * singular + if float(np.sum(energy)) <= 1e-15: + explained = np.zeros_like(energy) + effective_rank = 0.0 + else: + explained = energy / np.sum(energy) + pos = explained[explained > 1e-15] + effective_rank = float(np.exp(-np.sum(pos * np.log(pos)))) + + centroids = {} + for name in PERT_NAMES: + vals = np.array([v for v, label in zip(vectors, labels) if label == name], dtype=float) + if len(vals): + centroids[name] = np.mean(vals, axis=0) + + cosine = {} + for i, a_name in enumerate(PERT_NAMES): + for b_name in PERT_NAMES[i + 1 :]: + if a_name not in centroids or b_name not in centroids: + continue + a = centroids[a_name] + b = centroids[b_name] + denom = np.linalg.norm(a) * np.linalg.norm(b) + cosine[f"{a_name}_vs_{b_name}"] = float(np.dot(a, b) / denom) if denom > 1e-15 else 0.0 + + return { + "explained_variance": [float(x) for x in explained], + "effective_rank": effective_rank, + "centroid_cosine": cosine, + "pc2": float(explained[1]) if len(explained) > 1 else 0.0, + } + + +def analyze_sequence( + gaps: np.ndarray, + alphas: list[float], + n_trials: int, + n_baseline: int, + z_min: float, + rng: np.random.Generator, +) -> dict: + original = compute_canonical(gaps) + baseline_vals = {obs: [] for obs in OBS_NAMES} + for _ in range(n_baseline): + obs = compute_canonical(rng.permutation(gaps)) + for name in OBS_NAMES: + baseline_vals[name].append(obs[name]) + + baseline = {} + z = {} + denom = {} + for name in OBS_NAMES: + vals = np.array(baseline_vals[name], dtype=float) + mean = float(np.mean(vals)) + std = float(np.std(vals, ddof=1)) if len(vals) > 1 else 0.0 + baseline[name] = {"mean": mean, "std": std} + denom[name] = float(original[name] - mean) + z[name] = float(denom[name] / std) if std > 1e-15 else 0.0 + + stable_obs = [name for name in OBS_NAMES if abs(z[name]) >= z_min] + all_vectors = [] + screened_vectors = [] + labels = [] + profiles = [] + + for pert_name in PERT_NAMES: + for alpha in alphas: + trial_vals = {obs: [] for obs in OBS_NAMES} + for _ in range(n_trials): + perturbed = PERTURB[pert_name](gaps, alpha, rng) + obs = compute_canonical(perturbed) + for name in OBS_NAMES: + trial_vals[name].append(obs[name]) + means = {name: float(np.mean(trial_vals[name])) for name in OBS_NAMES} + retention = {} + for name in OBS_NAMES: + retention[name] = ( + float((means[name] - baseline[name]["mean"]) / denom[name]) + if abs(denom[name]) > 1e-12 + else 0.0 + ) + all_vector = [retention[name] for name in OBS_NAMES] + screened_vector = [retention[name] for name in stable_obs] + all_vectors.append(all_vector) + if len(stable_obs) >= 2: + screened_vectors.append(screened_vector) + labels.append(pert_name) + profiles.append( + { + "perturbation": pert_name, + "alpha": float(alpha), + "retention": retention, + "retention_vector": all_vector, + } + ) + + all_pca = pca_summary(all_vectors, labels) + screened_pca = pca_summary(screened_vectors, labels) if len(stable_obs) >= 2 else None + + return { + "n_gaps": int(len(gaps)), + "original": original, + "full_shuffle_baseline": baseline, + "denominator": denom, + "original_vs_shuffle_z": z, + "stable_observables": stable_obs, + "weak_observable_count": int(len(OBS_NAMES) - len(stable_obs)), + "profiles": profiles, + "pca_all_observables": all_pca, + "pca_stable_observables": screened_pca, + } + + +def summarize_replicates(items: list[dict]) -> dict: + def arr(path: tuple[str, ...]) -> np.ndarray: + vals = [] + for item in items: + x = item + for key in path: + if x is None: + break + x = x.get(key) + if isinstance(x, (int, float)): + vals.append(float(x)) + return np.array(vals, dtype=float) + + rank = arr(("pca_all_observables", "effective_rank")) + pc2 = arr(("pca_all_observables", "pc2")) + weak = np.array([item["weak_observable_count"] for item in items], dtype=float) + stable_rank = arr(("pca_stable_observables", "effective_rank")) + cos = arr(("pca_all_observables", "centroid_cosine", "adjacent_swap_vs_large_gap_only")) + + out = { + "n_replicates": len(items), + "rank_mean": float(np.mean(rank)) if len(rank) else 0.0, + "rank_std": float(np.std(rank, ddof=1)) if len(rank) > 1 else 0.0, + "pc2_mean": float(np.mean(pc2)) if len(pc2) else 0.0, + "pc2_std": float(np.std(pc2, ddof=1)) if len(pc2) > 1 else 0.0, + "weak_observable_count_mean": float(np.mean(weak)) if len(weak) else 0.0, + "stable_rank_mean": float(np.mean(stable_rank)) if len(stable_rank) else None, + "stable_rank_std": float(np.std(stable_rank, ddof=1)) if len(stable_rank) > 1 else 0.0, + "adjacent_vs_large_cosine_mean": float(np.mean(cos)) if len(cos) else 0.0, + "adjacent_vs_large_cosine_std": float(np.std(cos, ddof=1)) if len(cos) > 1 else 0.0, + } + if len(rank) > 1 and np.std(weak) > 1e-15 and np.std(rank) > 1e-15: + out["rank_vs_weak_count_corr"] = float(np.corrcoef(rank, weak)[0, 1]) + else: + out["rank_vs_weak_count_corr"] = 0.0 + return out + + +def build_prime_windows(max_n: int, n_reps: int) -> list[np.ndarray]: + total = max_n * n_reps + max_n + gaps = prime_gaps(total) + max_start = len(gaps) - max_n + starts = np.linspace(0, max_start, n_reps, dtype=int) + return [gaps[start : start + max_n].astype(float) for start in starts] + + +def run(args: argparse.Namespace) -> dict: + root_rng = np.random.default_rng(args.seed) + sizes = [int(x) for x in args.sizes.split(",") if x.strip()] + max_n = max(sizes) + alphas = [float(x) for x in np.linspace(args.alpha_min, args.alpha_max, args.n_alpha)] + + output = { + "experiment": "perturbation_rank_size_curve", + "question": "Does perturbation effective-rank stabilize with sample size under canonical observables?", + "observables_registry": OBSERVABLES_REGISTRY_VERSION, + "observables_used": OBS_NAMES, + "perturbations": PERT_NAMES, + "params": vars(args), + "alphas": alphas, + "domains": {}, + "summary": {}, + } + + prime_windows = build_prime_windows(max_n, args.n_replicates) + + print(f"observables_registry={OBSERVABLES_REGISTRY_VERSION}") + print(f"{'domain':<22} {'N':>6} {'rank':>7} {'PC2':>7} {'weak':>5} {'stable_rank':>11}") + + domain_builders = { + "primes_windows": lambda rep_rng, rep_i: prime_windows[rep_i], + "prime_shuffle_control": lambda rep_rng, rep_i: rep_rng.permutation(prime_windows[rep_i]), + "poisson": lambda rep_rng, rep_i: rep_rng.exponential(1.0, size=max_n), + "gue": lambda rep_rng, rep_i: gue_spacings(args.gue_matrix_size, max_n, rep_rng), + } + + for domain_name, builder in domain_builders.items(): + output["domains"][domain_name] = {} + output["summary"][domain_name] = {} + bases = [] + for rep_i in range(args.n_replicates): + rep_rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) + bases.append(builder(rep_rng, rep_i)) + + for n in sizes: + rows = [] + for rep_i, base in enumerate(bases): + rep_rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) + res = analyze_sequence( + base[:n], + alphas=alphas, + n_trials=args.n_trials, + n_baseline=args.n_baseline, + z_min=args.z_min, + rng=rep_rng, + ) + res["replicate"] = rep_i + rows.append(res) + summary = summarize_replicates(rows) + output["domains"][domain_name][str(n)] = rows + output["summary"][domain_name][str(n)] = summary + stable_rank = summary["stable_rank_mean"] + stable_text = f"{stable_rank:>11.3f}" if stable_rank is not None else f"{'NA':>11}" + print( + f"{domain_name:<22} {n:>6} {summary['rank_mean']:>7.3f} " + f"{summary['pc2_mean']:>7.3f} {summary['weak_observable_count_mean']:>5.2f} " + f"{stable_text}" + ) + + out_path = Path(args.out) + out_path.parent.mkdir(parents=True, exist_ok=True) + with out_path.open("w") as f: + json.dump(output, f, indent=2) + print(f"saved {out_path}") + return output + + +def main() -> None: + parser = argparse.ArgumentParser() + parser.add_argument("--sizes", default="128,256,512,1024,2048") + parser.add_argument("--n-replicates", type=int, default=8) + parser.add_argument("--gue-matrix-size", type=int, default=180) + parser.add_argument("--n-alpha", type=int, default=5) + parser.add_argument("--alpha-min", type=float, default=0.1) + parser.add_argument("--alpha-max", type=float, default=0.9) + parser.add_argument("--n-trials", type=int, default=8) + parser.add_argument("--n-baseline", type=int, default=16) + parser.add_argument("--z-min", type=float, default=2.0) + parser.add_argument("--seed", type=int, default=20260506) + parser.add_argument("--out", default="tools/data/perturbation_rank_size_curve.json") + args = parser.parse_args() + run(args) + + +if __name__ == "__main__": + main() exec /bin/bash -lc 'tail -n 80 tools/data/seme.json' in /opt/MM_D-ND succeeded in 0ms: "fluctuations", "information theory", "probability", "Boltzmann" ], "content": "Computing termodinamico: sfruttare le fluttuazioni termiche invece di combatterle. Il rumore non è nemico — è risorsa. Extropic costruisce hardware che usa l'entropia come motore computazionale. Connessione profonda tra termodinamica, teoria dell'informazione e probabilità.", "teorie": [ "T", "Q" ], "ponti_potenziali": [ { "coppia": "TxQ", "ponte": "noise come risorsa computazionale — fluttuazioni termiche = calcolo", "nota": "il vuoto quantistico (pieno di fluttuazioni) è il computer" } ], "timestamp": "2026-04-02T08:23:13.992019" }, { "id": "video_j0wJBEZdwLs", "title": "What is a Laplace Transform - visual explanation", "type": "video_digest", "keywords": [ "Laplace transform", "frequency", "damping", "s-plane", "complex", "exponential" ], "content": "La trasformata di Laplace come proiezione su esponenziali complesse. Il piano s = σ + iω combina smorzamento (reale) e oscillazione (immaginario). Connessione tra dominio temporale e dominio delle frequenze complesse.", "teorie": [ "T", "Q", "R" ], "ponti_potenziali": [], "timestamp": "2026-04-02T08:23:13.992021" }, { "id": "video_rZ2m1_q9lg0", "title": "New duality: conductor-insulator in YbB12 at 35T - University of Michigan", "type": "video_digest", "keywords": [ "duality", "conductor", "insulator", "Kondo insulator", "quantum oscillations", "ytterbium boride", "YbB12", "charge-neutral fermions", "strongly correlated", "condensed matter", "magnetic field" ], "content": "Scoperta dualita conduttore-isolante in YbB12 a 35 Tesla. Il materiale si comporta simultaneamente da conduttore e isolante — oscillazioni quantistiche dal bulk, non dalla superficie. Charge-neutral fermions (portatori senza carica) creano comportamento di conduzione = il terzo incluso. La scoperta avviene a campo estremo dove il confine tra stati collassa.", "teorie": [ "T", "Q", "E" ], "ponti_potenziali": [ { "coppia": "TxQ", "ponte": "Kondo insulator fortemente correlato — dualita conduttore/isolante come stato quantistico unico", "nota": "charge-neutral fermions = portatori senza carica che conducono, terzo incluso operativo" }, { "coppia": "ExQ", "ponte": "campo magnetico estremo (35T) collassa il confine tra stati — condizione di soglia", "nota": "il confine non separa, unifica. Pattern BOUNDARY del Domandatore" } ], "ref": "Physical Review Letters 2025, 135(15). Chen, Zhu, Ratkovski et al. PI: Lu Li", "timestamp": "2026-04-04T19:34:08.014Z" } ] } diff --git a/tools/exp_perturbation_rank_size_curve.py b/tools/exp_perturbation_rank_size_curve.py new file mode 100644 index 0000000000000000000000000000000000000000..032847ca3d82ec3233ec2b41ad230dfe394d2791 --- /dev/null +++ b/tools/exp_perturbation_rank_size_curve.py @@ -0,0 +1,387 @@ +#!/usr/bin/env python3 +""" +exp_perturbation_rank_size_curve.py + +Reusable META audit for perturbation dimensionality. + +The 2026-05-06 06:25 cycle restricted the claim "GUE has a second +perturbation axis" to sample size, generator, and observable definitions. +This tool measures the size curve directly, using the canonical observable +registry and explicit original-vs-shuffle denominator diagnostics. + +The report owns interpretation. This script only measures: +- effective rank and PC2 across scale-selective perturbation profiles; +- original-vs-shuffle z-score per observable; +- whether apparent rank co-occurs with weak retention denominators. +""" + +from __future__ import annotations + +import argparse +import json +from pathlib import Path + +import numpy as np + +from observables_registry import ( + OBSERVABLES_CANONICAL, + OBSERVABLES_REGISTRY_VERSION, + compute_canonical, +) + + +OBS_NAMES = list(OBSERVABLES_CANONICAL.keys()) +PERT_NAMES = ["adjacent_swap", "block_shuffle", "large_gap_only", "uniform"] + + +def prime_gaps(n_gaps: int) -> np.ndarray: + """Return the first n_gaps prime gaps using a compact numpy sieve.""" + limit = max(100, int(n_gaps * (np.log(n_gaps + 10) + np.log(np.log(n_gaps + 10)) + 5))) + while True: + sieve = np.ones(limit + 1, dtype=bool) + sieve[:2] = False + for p in range(2, int(limit**0.5) + 1): + if sieve[p]: + sieve[p * p : limit + 1 : p] = False + primes = np.flatnonzero(sieve) + if len(primes) >= n_gaps + 1: + return np.diff(primes[: n_gaps + 1]).astype(float) + limit *= 2 + + +def gue_spacings(matrix_size: int, min_spacings: int, rng: np.random.Generator) -> np.ndarray: + """Generate unfolded GUE spacings by concatenating independent matrices.""" + parts = [] + edge = max(2, matrix_size // 10) + while sum(len(x) for x in parts) < min_spacings: + real = rng.standard_normal((matrix_size, matrix_size)) + imag = rng.standard_normal((matrix_size, matrix_size)) + h = real + 1j * imag + h = (h + h.conj().T) / (2.0 * np.sqrt(matrix_size)) + eigs = np.sort(np.linalg.eigvalsh(h).real) + bulk = eigs[edge:-edge] + gaps = np.diff(bulk) + mean = float(np.mean(gaps)) + if mean > 1e-15: + parts.append(gaps / mean) + return np.concatenate(parts)[:min_spacings].astype(float) + + +def perturb_adjacent_swap(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + idx = np.arange(0, len(out) - 1, 2) + chosen = idx[rng.random(len(idx)) < alpha] + tmp = out[chosen].copy() + out[chosen] = out[chosen + 1] + out[chosen + 1] = tmp + return out + + +def perturb_block_shuffle(gaps: np.ndarray, alpha: float, rng: np.random.Generator, block_size: int = 64) -> np.ndarray: + out = gaps.copy() + n_blocks = len(out) // block_size + if n_blocks <= 0: + return rng.permutation(out) if alpha > 0 else out + k = int(round(alpha * n_blocks)) + if k <= 0: + return out + for block in rng.choice(n_blocks, size=min(k, n_blocks), replace=False): + start = block * block_size + end = min(start + block_size, len(out)) + rng.shuffle(out[start:end]) + return out + + +def perturb_large_gap_only(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + idx = np.flatnonzero(out > np.median(out)) + k = int(round(alpha * len(idx))) + if k < 2: + return out + chosen = rng.choice(idx, size=min(k, len(idx)), replace=False) + vals = out[chosen].copy() + rng.shuffle(vals) + out[chosen] = vals + return out + + +def perturb_uniform(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + k = int(round(alpha * len(out))) + if k < 2: + return out + chosen = rng.choice(len(out), size=min(k, len(out)), replace=False) + vals = out[chosen].copy() + rng.shuffle(vals) + out[chosen] = vals + return out + + +PERTURB = { + "adjacent_swap": perturb_adjacent_swap, + "block_shuffle": perturb_block_shuffle, + "large_gap_only": perturb_large_gap_only, + "uniform": perturb_uniform, +} + + +def pca_summary(vectors: list[list[float]], labels: list[str]) -> dict: + matrix = np.array(vectors, dtype=float) + matrix = matrix - np.mean(matrix, axis=0, keepdims=True) + if min(matrix.shape) == 0 or np.max(np.abs(matrix)) <= 1e-15: + return { + "explained_variance": [], + "effective_rank": 0.0, + "centroid_cosine": {}, + "pc2": 0.0, + } + _, singular, _ = np.linalg.svd(matrix, full_matrices=False) + energy = singular * singular + if float(np.sum(energy)) <= 1e-15: + explained = np.zeros_like(energy) + effective_rank = 0.0 + else: + explained = energy / np.sum(energy) + pos = explained[explained > 1e-15] + effective_rank = float(np.exp(-np.sum(pos * np.log(pos)))) + + centroids = {} + for name in PERT_NAMES: + vals = np.array([v for v, label in zip(vectors, labels) if label == name], dtype=float) + if len(vals): + centroids[name] = np.mean(vals, axis=0) + + cosine = {} + for i, a_name in enumerate(PERT_NAMES): + for b_name in PERT_NAMES[i + 1 :]: + if a_name not in centroids or b_name not in centroids: + continue + a = centroids[a_name] + b = centroids[b_name] + denom = np.linalg.norm(a) * np.linalg.norm(b) + cosine[f"{a_name}_vs_{b_name}"] = float(np.dot(a, b) / denom) if denom > 1e-15 else 0.0 + + return { + "explained_variance": [float(x) for x in explained], + "effective_rank": effective_rank, + "centroid_cosine": cosine, + "pc2": float(explained[1]) if len(explained) > 1 else 0.0, + } + + +def analyze_sequence( + gaps: np.ndarray, + alphas: list[float], + n_trials: int, + n_baseline: int, + z_min: float, + rng: np.random.Generator, +) -> dict: + original = compute_canonical(gaps) + baseline_vals = {obs: [] for obs in OBS_NAMES} + for _ in range(n_baseline): + obs = compute_canonical(rng.permutation(gaps)) + for name in OBS_NAMES: + baseline_vals[name].append(obs[name]) + + baseline = {} + z = {} + denom = {} + for name in OBS_NAMES: + vals = np.array(baseline_vals[name], dtype=float) + mean = float(np.mean(vals)) + std = float(np.std(vals, ddof=1)) if len(vals) > 1 else 0.0 + baseline[name] = {"mean": mean, "std": std} + denom[name] = float(original[name] - mean) + z[name] = float(denom[name] / std) if std > 1e-15 else 0.0 + + stable_obs = [name for name in OBS_NAMES if abs(z[name]) >= z_min] + all_vectors = [] + screened_vectors = [] + labels = [] + profiles = [] + + for pert_name in PERT_NAMES: + for alpha in alphas: + trial_vals = {obs: [] for obs in OBS_NAMES} + for _ in range(n_trials): + perturbed = PERTURB[pert_name](gaps, alpha, rng) + obs = compute_canonical(perturbed) + for name in OBS_NAMES: + trial_vals[name].append(obs[name]) + means = {name: float(np.mean(trial_vals[name])) for name in OBS_NAMES} + retention = {} + for name in OBS_NAMES: + retention[name] = ( + float((means[name] - baseline[name]["mean"]) / denom[name]) + if abs(denom[name]) > 1e-12 + else 0.0 + ) + all_vector = [retention[name] for name in OBS_NAMES] + screened_vector = [retention[name] for name in stable_obs] + all_vectors.append(all_vector) + if len(stable_obs) >= 2: + screened_vectors.append(screened_vector) + labels.append(pert_name) + profiles.append( + { + "perturbation": pert_name, + "alpha": float(alpha), + "retention": retention, + "retention_vector": all_vector, + } + ) + + all_pca = pca_summary(all_vectors, labels) + screened_pca = pca_summary(screened_vectors, labels) if len(stable_obs) >= 2 else None + + return { + "n_gaps": int(len(gaps)), + "original": original, + "full_shuffle_baseline": baseline, + "denominator": denom, + "original_vs_shuffle_z": z, + "stable_observables": stable_obs, + "weak_observable_count": int(len(OBS_NAMES) - len(stable_obs)), + "profiles": profiles, + "pca_all_observables": all_pca, + "pca_stable_observables": screened_pca, + } + + +def summarize_replicates(items: list[dict]) -> dict: + def arr(path: tuple[str, ...]) -> np.ndarray: + vals = [] + for item in items: + x = item + for key in path: + if x is None: + break + x = x.get(key) + if isinstance(x, (int, float)): + vals.append(float(x)) + return np.array(vals, dtype=float) + + rank = arr(("pca_all_observables", "effective_rank")) + pc2 = arr(("pca_all_observables", "pc2")) + weak = np.array([item["weak_observable_count"] for item in items], dtype=float) + stable_rank = arr(("pca_stable_observables", "effective_rank")) + cos = arr(("pca_all_observables", "centroid_cosine", "adjacent_swap_vs_large_gap_only")) + + out = { + "n_replicates": len(items), + "rank_mean": float(np.mean(rank)) if len(rank) else 0.0, + "rank_std": float(np.std(rank, ddof=1)) if len(rank) > 1 else 0.0, + "pc2_mean": float(np.mean(pc2)) if len(pc2) else 0.0, + "pc2_std": float(np.std(pc2, ddof=1)) if len(pc2) > 1 else 0.0, + "weak_observable_count_mean": float(np.mean(weak)) if len(weak) else 0.0, + "stable_rank_mean": float(np.mean(stable_rank)) if len(stable_rank) else None, + "stable_rank_std": float(np.std(stable_rank, ddof=1)) if len(stable_rank) > 1 else 0.0, + "adjacent_vs_large_cosine_mean": float(np.mean(cos)) if len(cos) else 0.0, + "adjacent_vs_large_cosine_std": float(np.std(cos, ddof=1)) if len(cos) > 1 else 0.0, + } + if len(rank) > 1 and np.std(weak) > 1e-15 and np.std(rank) > 1e-15: + out["rank_vs_weak_count_corr"] = float(np.corrcoef(rank, weak)[0, 1]) + else: + out["rank_vs_weak_count_corr"] = 0.0 + return out + + +def build_prime_windows(max_n: int, n_reps: int) -> list[np.ndarray]: + total = max_n * n_reps + max_n + gaps = prime_gaps(total) + max_start = len(gaps) - max_n + starts = np.linspace(0, max_start, n_reps, dtype=int) + return [gaps[start : start + max_n].astype(float) for start in starts] + + +def run(args: argparse.Namespace) -> dict: + root_rng = np.random.default_rng(args.seed) + sizes = [int(x) for x in args.sizes.split(",") if x.strip()] + max_n = max(sizes) + alphas = [float(x) for x in np.linspace(args.alpha_min, args.alpha_max, args.n_alpha)] + + output = { + "experiment": "perturbation_rank_size_curve", + "question": "Does perturbation effective-rank stabilize with sample size under canonical observables?", + "observables_registry": OBSERVABLES_REGISTRY_VERSION, + "observables_used": OBS_NAMES, + "perturbations": PERT_NAMES, + "params": vars(args), + "alphas": alphas, + "domains": {}, + "summary": {}, + } + + prime_windows = build_prime_windows(max_n, args.n_replicates) + + print(f"observables_registry={OBSERVABLES_REGISTRY_VERSION}") + print(f"{'domain':<22} {'N':>6} {'rank':>7} {'PC2':>7} {'weak':>5} {'stable_rank':>11}") + + domain_builders = { + "primes_windows": lambda rep_rng, rep_i: prime_windows[rep_i], + "prime_shuffle_control": lambda rep_rng, rep_i: rep_rng.permutation(prime_windows[rep_i]), + "poisson": lambda rep_rng, rep_i: rep_rng.exponential(1.0, size=max_n), + "gue": lambda rep_rng, rep_i: gue_spacings(args.gue_matrix_size, max_n, rep_rng), + } + + for domain_name, builder in domain_builders.items(): + output["domains"][domain_name] = {} + output["summary"][domain_name] = {} + bases = [] + for rep_i in range(args.n_replicates): + rep_rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) + bases.append(builder(rep_rng, rep_i)) + + for n in sizes: + rows = [] + for rep_i, base in enumerate(bases): + rep_rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) + res = analyze_sequence( + base[:n], + alphas=alphas, + n_trials=args.n_trials, + n_baseline=args.n_baseline, + z_min=args.z_min, + rng=rep_rng, + ) + res["replicate"] = rep_i + rows.append(res) + summary = summarize_replicates(rows) + output["domains"][domain_name][str(n)] = rows + output["summary"][domain_name][str(n)] = summary + stable_rank = summary["stable_rank_mean"] + stable_text = f"{stable_rank:>11.3f}" if stable_rank is not None else f"{'NA':>11}" + print( + f"{domain_name:<22} {n:>6} {summary['rank_mean']:>7.3f} " + f"{summary['pc2_mean']:>7.3f} {summary['weak_observable_count_mean']:>5.2f} " + f"{stable_text}" + ) + + out_path = Path(args.out) + out_path.parent.mkdir(parents=True, exist_ok=True) + with out_path.open("w") as f: + json.dump(output, f, indent=2) + print(f"saved {out_path}") + return output + + +def main() -> None: + parser = argparse.ArgumentParser() + parser.add_argument("--sizes", default="128,256,512,1024,2048") + parser.add_argument("--n-replicates", type=int, default=8) + parser.add_argument("--gue-matrix-size", type=int, default=180) + parser.add_argument("--n-alpha", type=int, default=5) + parser.add_argument("--alpha-min", type=float, default=0.1) + parser.add_argument("--alpha-max", type=float, default=0.9) + parser.add_argument("--n-trials", type=int, default=8) + parser.add_argument("--n-baseline", type=int, default=16) + parser.add_argument("--z-min", type=float, default=2.0) + parser.add_argument("--seed", type=int, default=20260506) + parser.add_argument("--out", default="tools/data/perturbation_rank_size_curve.json") + args = parser.parse_args() + run(args) + + +if __name__ == "__main__": + main() exec /bin/bash -lc 'rg -n "OBSERVABLE_REGISTRY|fonti_consumate" tools/data/seme.json' in /opt/MM_D-ND succeeded in 0ms: 92: "id": "OBSERVABLE_REGISTRY", 119: "fonti_consumate": 0, diff --git a/tools/exp_perturbation_rank_size_curve.py b/tools/exp_perturbation_rank_size_curve.py new file mode 100644 index 0000000000000000000000000000000000000000..032847ca3d82ec3233ec2b41ad230dfe394d2791 --- /dev/null +++ b/tools/exp_perturbation_rank_size_curve.py @@ -0,0 +1,387 @@ +#!/usr/bin/env python3 +""" +exp_perturbation_rank_size_curve.py + +Reusable META audit for perturbation dimensionality. + +The 2026-05-06 06:25 cycle restricted the claim "GUE has a second +perturbation axis" to sample size, generator, and observable definitions. +This tool measures the size curve directly, using the canonical observable +registry and explicit original-vs-shuffle denominator diagnostics. + +The report owns interpretation. This script only measures: +- effective rank and PC2 across scale-selective perturbation profiles; +- original-vs-shuffle z-score per observable; +- whether apparent rank co-occurs with weak retention denominators. +""" + +from __future__ import annotations + +import argparse +import json +from pathlib import Path + +import numpy as np + +from observables_registry import ( + OBSERVABLES_CANONICAL, + OBSERVABLES_REGISTRY_VERSION, + compute_canonical, +) + + +OBS_NAMES = list(OBSERVABLES_CANONICAL.keys()) +PERT_NAMES = ["adjacent_swap", "block_shuffle", "large_gap_only", "uniform"] + + +def prime_gaps(n_gaps: int) -> np.ndarray: + """Return the first n_gaps prime gaps using a compact numpy sieve.""" + limit = max(100, int(n_gaps * (np.log(n_gaps + 10) + np.log(np.log(n_gaps + 10)) + 5))) + while True: + sieve = np.ones(limit + 1, dtype=bool) + sieve[:2] = False + for p in range(2, int(limit**0.5) + 1): + if sieve[p]: + sieve[p * p : limit + 1 : p] = False + primes = np.flatnonzero(sieve) + if len(primes) >= n_gaps + 1: + return np.diff(primes[: n_gaps + 1]).astype(float) + limit *= 2 + + +def gue_spacings(matrix_size: int, min_spacings: int, rng: np.random.Generator) -> np.ndarray: + """Generate unfolded GUE spacings by concatenating independent matrices.""" + parts = [] + edge = max(2, matrix_size // 10) + while sum(len(x) for x in parts) < min_spacings: + real = rng.standard_normal((matrix_size, matrix_size)) + imag = rng.standard_normal((matrix_size, matrix_size)) + h = real + 1j * imag + h = (h + h.conj().T) / (2.0 * np.sqrt(matrix_size)) + eigs = np.sort(np.linalg.eigvalsh(h).real) + bulk = eigs[edge:-edge] + gaps = np.diff(bulk) + mean = float(np.mean(gaps)) + if mean > 1e-15: + parts.append(gaps / mean) + return np.concatenate(parts)[:min_spacings].astype(float) + + +def perturb_adjacent_swap(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + idx = np.arange(0, len(out) - 1, 2) + chosen = idx[rng.random(len(idx)) < alpha] + tmp = out[chosen].copy() + out[chosen] = out[chosen + 1] + out[chosen + 1] = tmp + return out + + +def perturb_block_shuffle(gaps: np.ndarray, alpha: float, rng: np.random.Generator, block_size: int = 64) -> np.ndarray: + out = gaps.copy() + n_blocks = len(out) // block_size + if n_blocks <= 0: + return rng.permutation(out) if alpha > 0 else out + k = int(round(alpha * n_blocks)) + if k <= 0: + return out + for block in rng.choice(n_blocks, size=min(k, n_blocks), replace=False): + start = block * block_size + end = min(start + block_size, len(out)) + rng.shuffle(out[start:end]) + return out + + +def perturb_large_gap_only(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + idx = np.flatnonzero(out > np.median(out)) + k = int(round(alpha * len(idx))) + if k < 2: + return out + chosen = rng.choice(idx, size=min(k, len(idx)), replace=False) + vals = out[chosen].copy() + rng.shuffle(vals) + out[chosen] = vals + return out + + +def perturb_uniform(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + k = int(round(alpha * len(out))) + if k < 2: + return out + chosen = rng.choice(len(out), size=min(k, len(out)), replace=False) + vals = out[chosen].copy() + rng.shuffle(vals) + out[chosen] = vals + return out + + +PERTURB = { + "adjacent_swap": perturb_adjacent_swap, + "block_shuffle": perturb_block_shuffle, + "large_gap_only": perturb_large_gap_only, + "uniform": perturb_uniform, +} + + +def pca_summary(vectors: list[list[float]], labels: list[str]) -> dict: + matrix = np.array(vectors, dtype=float) + matrix = matrix - np.mean(matrix, axis=0, keepdims=True) + if min(matrix.shape) == 0 or np.max(np.abs(matrix)) <= 1e-15: + return { + "explained_variance": [], + "effective_rank": 0.0, + "centroid_cosine": {}, + "pc2": 0.0, + } + _, singular, _ = np.linalg.svd(matrix, full_matrices=False) + energy = singular * singular + if float(np.sum(energy)) <= 1e-15: + explained = np.zeros_like(energy) + effective_rank = 0.0 + else: + explained = energy / np.sum(energy) + pos = explained[explained > 1e-15] + effective_rank = float(np.exp(-np.sum(pos * np.log(pos)))) + + centroids = {} + for name in PERT_NAMES: + vals = np.array([v for v, label in zip(vectors, labels) if label == name], dtype=float) + if len(vals): + centroids[name] = np.mean(vals, axis=0) + + cosine = {} + for i, a_name in enumerate(PERT_NAMES): + for b_name in PERT_NAMES[i + 1 :]: + if a_name not in centroids or b_name not in centroids: + continue + a = centroids[a_name] + b = centroids[b_name] + denom = np.linalg.norm(a) * np.linalg.norm(b) + cosine[f"{a_name}_vs_{b_name}"] = float(np.dot(a, b) / denom) if denom > 1e-15 else 0.0 + + return { + "explained_variance": [float(x) for x in explained], + "effective_rank": effective_rank, + "centroid_cosine": cosine, + "pc2": float(explained[1]) if len(explained) > 1 else 0.0, + } + + +def analyze_sequence( + gaps: np.ndarray, + alphas: list[float], + n_trials: int, + n_baseline: int, + z_min: float, + rng: np.random.Generator, +) -> dict: + original = compute_canonical(gaps) + baseline_vals = {obs: [] for obs in OBS_NAMES} + for _ in range(n_baseline): + obs = compute_canonical(rng.permutation(gaps)) + for name in OBS_NAMES: + baseline_vals[name].append(obs[name]) + + baseline = {} + z = {} + denom = {} + for name in OBS_NAMES: + vals = np.array(baseline_vals[name], dtype=float) + mean = float(np.mean(vals)) + std = float(np.std(vals, ddof=1)) if len(vals) > 1 else 0.0 + baseline[name] = {"mean": mean, "std": std} + denom[name] = float(original[name] - mean) + z[name] = float(denom[name] / std) if std > 1e-15 else 0.0 + + stable_obs = [name for name in OBS_NAMES if abs(z[name]) >= z_min] + all_vectors = [] + screened_vectors = [] + labels = [] + profiles = [] + + for pert_name in PERT_NAMES: + for alpha in alphas: + trial_vals = {obs: [] for obs in OBS_NAMES} + for _ in range(n_trials): + perturbed = PERTURB[pert_name](gaps, alpha, rng) + obs = compute_canonical(perturbed) + for name in OBS_NAMES: + trial_vals[name].append(obs[name]) + means = {name: float(np.mean(trial_vals[name])) for name in OBS_NAMES} + retention = {} + for name in OBS_NAMES: + retention[name] = ( + float((means[name] - baseline[name]["mean"]) / denom[name]) + if abs(denom[name]) > 1e-12 + else 0.0 + ) + all_vector = [retention[name] for name in OBS_NAMES] + screened_vector = [retention[name] for name in stable_obs] + all_vectors.append(all_vector) + if len(stable_obs) >= 2: + screened_vectors.append(screened_vector) + labels.append(pert_name) + profiles.append( + { + "perturbation": pert_name, + "alpha": float(alpha), + "retention": retention, + "retention_vector": all_vector, + } + ) + + all_pca = pca_summary(all_vectors, labels) + screened_pca = pca_summary(screened_vectors, labels) if len(stable_obs) >= 2 else None + + return { + "n_gaps": int(len(gaps)), + "original": original, + "full_shuffle_baseline": baseline, + "denominator": denom, + "original_vs_shuffle_z": z, + "stable_observables": stable_obs, + "weak_observable_count": int(len(OBS_NAMES) - len(stable_obs)), + "profiles": profiles, + "pca_all_observables": all_pca, + "pca_stable_observables": screened_pca, + } + + +def summarize_replicates(items: list[dict]) -> dict: + def arr(path: tuple[str, ...]) -> np.ndarray: + vals = [] + for item in items: + x = item + for key in path: + if x is None: + break + x = x.get(key) + if isinstance(x, (int, float)): + vals.append(float(x)) + return np.array(vals, dtype=float) + + rank = arr(("pca_all_observables", "effective_rank")) + pc2 = arr(("pca_all_observables", "pc2")) + weak = np.array([item["weak_observable_count"] for item in items], dtype=float) + stable_rank = arr(("pca_stable_observables", "effective_rank")) + cos = arr(("pca_all_observables", "centroid_cosine", "adjacent_swap_vs_large_gap_only")) + + out = { + "n_replicates": len(items), + "rank_mean": float(np.mean(rank)) if len(rank) else 0.0, + "rank_std": float(np.std(rank, ddof=1)) if len(rank) > 1 else 0.0, + "pc2_mean": float(np.mean(pc2)) if len(pc2) else 0.0, + "pc2_std": float(np.std(pc2, ddof=1)) if len(pc2) > 1 else 0.0, + "weak_observable_count_mean": float(np.mean(weak)) if len(weak) else 0.0, + "stable_rank_mean": float(np.mean(stable_rank)) if len(stable_rank) else None, + "stable_rank_std": float(np.std(stable_rank, ddof=1)) if len(stable_rank) > 1 else 0.0, + "adjacent_vs_large_cosine_mean": float(np.mean(cos)) if len(cos) else 0.0, + "adjacent_vs_large_cosine_std": float(np.std(cos, ddof=1)) if len(cos) > 1 else 0.0, + } + if len(rank) > 1 and np.std(weak) > 1e-15 and np.std(rank) > 1e-15: + out["rank_vs_weak_count_corr"] = float(np.corrcoef(rank, weak)[0, 1]) + else: + out["rank_vs_weak_count_corr"] = 0.0 + return out + + +def build_prime_windows(max_n: int, n_reps: int) -> list[np.ndarray]: + total = max_n * n_reps + max_n + gaps = prime_gaps(total) + max_start = len(gaps) - max_n + starts = np.linspace(0, max_start, n_reps, dtype=int) + return [gaps[start : start + max_n].astype(float) for start in starts] + + +def run(args: argparse.Namespace) -> dict: + root_rng = np.random.default_rng(args.seed) + sizes = [int(x) for x in args.sizes.split(",") if x.strip()] + max_n = max(sizes) + alphas = [float(x) for x in np.linspace(args.alpha_min, args.alpha_max, args.n_alpha)] + + output = { + "experiment": "perturbation_rank_size_curve", + "question": "Does perturbation effective-rank stabilize with sample size under canonical observables?", + "observables_registry": OBSERVABLES_REGISTRY_VERSION, + "observables_used": OBS_NAMES, + "perturbations": PERT_NAMES, + "params": vars(args), + "alphas": alphas, + "domains": {}, + "summary": {}, + } + + prime_windows = build_prime_windows(max_n, args.n_replicates) + + print(f"observables_registry={OBSERVABLES_REGISTRY_VERSION}") + print(f"{'domain':<22} {'N':>6} {'rank':>7} {'PC2':>7} {'weak':>5} {'stable_rank':>11}") + + domain_builders = { + "primes_windows": lambda rep_rng, rep_i: prime_windows[rep_i], + "prime_shuffle_control": lambda rep_rng, rep_i: rep_rng.permutation(prime_windows[rep_i]), + "poisson": lambda rep_rng, rep_i: rep_rng.exponential(1.0, size=max_n), + "gue": lambda rep_rng, rep_i: gue_spacings(args.gue_matrix_size, max_n, rep_rng), + } + + for domain_name, builder in domain_builders.items(): + output["domains"][domain_name] = {} + output["summary"][domain_name] = {} + bases = [] + for rep_i in range(args.n_replicates): + rep_rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) + bases.append(builder(rep_rng, rep_i)) + + for n in sizes: + rows = [] + for rep_i, base in enumerate(bases): + rep_rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) + res = analyze_sequence( + base[:n], + alphas=alphas, + n_trials=args.n_trials, + n_baseline=args.n_baseline, + z_min=args.z_min, + rng=rep_rng, + ) + res["replicate"] = rep_i + rows.append(res) + summary = summarize_replicates(rows) + output["domains"][domain_name][str(n)] = rows + output["summary"][domain_name][str(n)] = summary + stable_rank = summary["stable_rank_mean"] + stable_text = f"{stable_rank:>11.3f}" if stable_rank is not None else f"{'NA':>11}" + print( + f"{domain_name:<22} {n:>6} {summary['rank_mean']:>7.3f} " + f"{summary['pc2_mean']:>7.3f} {summary['weak_observable_count_mean']:>5.2f} " + f"{stable_text}" + ) + + out_path = Path(args.out) + out_path.parent.mkdir(parents=True, exist_ok=True) + with out_path.open("w") as f: + json.dump(output, f, indent=2) + print(f"saved {out_path}") + return output + + +def main() -> None: + parser = argparse.ArgumentParser() + parser.add_argument("--sizes", default="128,256,512,1024,2048") + parser.add_argument("--n-replicates", type=int, default=8) + parser.add_argument("--gue-matrix-size", type=int, default=180) + parser.add_argument("--n-alpha", type=int, default=5) + parser.add_argument("--alpha-min", type=float, default=0.1) + parser.add_argument("--alpha-max", type=float, default=0.9) + parser.add_argument("--n-trials", type=int, default=8) + parser.add_argument("--n-baseline", type=int, default=16) + parser.add_argument("--z-min", type=float, default=2.0) + parser.add_argument("--seed", type=int, default=20260506) + parser.add_argument("--out", default="tools/data/perturbation_rank_size_curve.json") + args = parser.parse_args() + run(args) + + +if __name__ == "__main__": + main() exec /bin/bash -lc "sed -n '82,122p' tools/data/seme.json" in /opt/MM_D-ND succeeded in 0ms: "tipo": "simmetria_sospetta", "id": "META", "claim": "Tutti i 11 test passano — verifica che non stiamo testando solo tautologie", "intensità": 0.5, "nota": "La convergenza a φ è triviale (controprove). I test stanno verificando contenuto o struttura?", "condensato_ref": "A4,A12,C2", "porta": "verify_assertions_META_ALL_PASS", "condensato_motivo": "Ricorrente (3x in 2 giorni) e fuori dalla mappa" }, { "id": "OBSERVABLE_REGISTRY", "tipo": "vincolo", "claim": "Ogni script che usa observables canonici (SR, SR2, L1, L2, triple_var) deve importare la definizione da tools/observables_registry.py. Varianti devono usare nomi distinti (SR_local_rigidity, triple_var_normalized) — niente shadowing del nome canonico. Ogni report deve dichiarare 'observables_registry: VERSION' nel header.", "intensita": 1.0, "porta": "infrastructure", "manuale": true, "condensato_ref": "A14,A8", "origine": "cristallizzato 06/05 dalla consecutio del cycle 20260506_0625 (autopoietico self-finding)", "added_at": "2026-05-06T07:03:58.213606+00:00" } ], "potenziale_bloccato": [], "varianza": [ "Tensioni risolte: {'PIANO_PRIMARIO_DUE_ASSIOMI', 'TENSIONE_ENTITA', 'TRASCENDENZA_LIMITE', 'METRIC_TENSOR', 'DUALITA_DIPOLARE_VS_ILLUSORIA', 'G_POTENZIALE_NULLA'}" ], "filtro": { "promosse": 8, "filtrate": 0, "condensato_claims": 30 }, "direzione": "Esplorare il confine: 8 domini GUE, 5 Poisson — il confine è il terzo incluso operativo", "verifica": { "pass": 11, "fail": 0, "skip": 0, "total": 11 }, "fonti_consumate": 0, "fonti_esterne": [ { "id": "video_lp0RgZ6kQF8", diff --git a/tools/exp_perturbation_rank_size_curve.py b/tools/exp_perturbation_rank_size_curve.py new file mode 100644 index 0000000000000000000000000000000000000000..032847ca3d82ec3233ec2b41ad230dfe394d2791 --- /dev/null +++ b/tools/exp_perturbation_rank_size_curve.py @@ -0,0 +1,387 @@ +#!/usr/bin/env python3 +""" +exp_perturbation_rank_size_curve.py + +Reusable META audit for perturbation dimensionality. + +The 2026-05-06 06:25 cycle restricted the claim "GUE has a second +perturbation axis" to sample size, generator, and observable definitions. +This tool measures the size curve directly, using the canonical observable +registry and explicit original-vs-shuffle denominator diagnostics. + +The report owns interpretation. This script only measures: +- effective rank and PC2 across scale-selective perturbation profiles; +- original-vs-shuffle z-score per observable; +- whether apparent rank co-occurs with weak retention denominators. +""" + +from __future__ import annotations + +import argparse +import json +from pathlib import Path + +import numpy as np + +from observables_registry import ( + OBSERVABLES_CANONICAL, + OBSERVABLES_REGISTRY_VERSION, + compute_canonical, +) + + +OBS_NAMES = list(OBSERVABLES_CANONICAL.keys()) +PERT_NAMES = ["adjacent_swap", "block_shuffle", "large_gap_only", "uniform"] + + +def prime_gaps(n_gaps: int) -> np.ndarray: + """Return the first n_gaps prime gaps using a compact numpy sieve.""" + limit = max(100, int(n_gaps * (np.log(n_gaps + 10) + np.log(np.log(n_gaps + 10)) + 5))) + while True: + sieve = np.ones(limit + 1, dtype=bool) + sieve[:2] = False + for p in range(2, int(limit**0.5) + 1): + if sieve[p]: + sieve[p * p : limit + 1 : p] = False + primes = np.flatnonzero(sieve) + if len(primes) >= n_gaps + 1: + return np.diff(primes[: n_gaps + 1]).astype(float) + limit *= 2 + + +def gue_spacings(matrix_size: int, min_spacings: int, rng: np.random.Generator) -> np.ndarray: + """Generate unfolded GUE spacings by concatenating independent matrices.""" + parts = [] + edge = max(2, matrix_size // 10) + while sum(len(x) for x in parts) < min_spacings: + real = rng.standard_normal((matrix_size, matrix_size)) + imag = rng.standard_normal((matrix_size, matrix_size)) + h = real + 1j * imag + h = (h + h.conj().T) / (2.0 * np.sqrt(matrix_size)) + eigs = np.sort(np.linalg.eigvalsh(h).real) + bulk = eigs[edge:-edge] + gaps = np.diff(bulk) + mean = float(np.mean(gaps)) + if mean > 1e-15: + parts.append(gaps / mean) + return np.concatenate(parts)[:min_spacings].astype(float) + + +def perturb_adjacent_swap(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + idx = np.arange(0, len(out) - 1, 2) + chosen = idx[rng.random(len(idx)) < alpha] + tmp = out[chosen].copy() + out[chosen] = out[chosen + 1] + out[chosen + 1] = tmp + return out + + +def perturb_block_shuffle(gaps: np.ndarray, alpha: float, rng: np.random.Generator, block_size: int = 64) -> np.ndarray: + out = gaps.copy() + n_blocks = len(out) // block_size + if n_blocks <= 0: + return rng.permutation(out) if alpha > 0 else out + k = int(round(alpha * n_blocks)) + if k <= 0: + return out + for block in rng.choice(n_blocks, size=min(k, n_blocks), replace=False): + start = block * block_size + end = min(start + block_size, len(out)) + rng.shuffle(out[start:end]) + return out + + +def perturb_large_gap_only(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + idx = np.flatnonzero(out > np.median(out)) + k = int(round(alpha * len(idx))) + if k < 2: + return out + chosen = rng.choice(idx, size=min(k, len(idx)), replace=False) + vals = out[chosen].copy() + rng.shuffle(vals) + out[chosen] = vals + return out + + +def perturb_uniform(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + k = int(round(alpha * len(out))) + if k < 2: + return out + chosen = rng.choice(len(out), size=min(k, len(out)), replace=False) + vals = out[chosen].copy() + rng.shuffle(vals) + out[chosen] = vals + return out + + +PERTURB = { + "adjacent_swap": perturb_adjacent_swap, + "block_shuffle": perturb_block_shuffle, + "large_gap_only": perturb_large_gap_only, + "uniform": perturb_uniform, +} + + +def pca_summary(vectors: list[list[float]], labels: list[str]) -> dict: + matrix = np.array(vectors, dtype=float) + matrix = matrix - np.mean(matrix, axis=0, keepdims=True) + if min(matrix.shape) == 0 or np.max(np.abs(matrix)) <= 1e-15: + return { + "explained_variance": [], + "effective_rank": 0.0, + "centroid_cosine": {}, + "pc2": 0.0, + } + _, singular, _ = np.linalg.svd(matrix, full_matrices=False) + energy = singular * singular + if float(np.sum(energy)) <= 1e-15: + explained = np.zeros_like(energy) + effective_rank = 0.0 + else: + explained = energy / np.sum(energy) + pos = explained[explained > 1e-15] + effective_rank = float(np.exp(-np.sum(pos * np.log(pos)))) + + centroids = {} + for name in PERT_NAMES: + vals = np.array([v for v, label in zip(vectors, labels) if label == name], dtype=float) + if len(vals): + centroids[name] = np.mean(vals, axis=0) + + cosine = {} + for i, a_name in enumerate(PERT_NAMES): + for b_name in PERT_NAMES[i + 1 :]: + if a_name not in centroids or b_name not in centroids: + continue + a = centroids[a_name] + b = centroids[b_name] + denom = np.linalg.norm(a) * np.linalg.norm(b) + cosine[f"{a_name}_vs_{b_name}"] = float(np.dot(a, b) / denom) if denom > 1e-15 else 0.0 + + return { + "explained_variance": [float(x) for x in explained], + "effective_rank": effective_rank, + "centroid_cosine": cosine, + "pc2": float(explained[1]) if len(explained) > 1 else 0.0, + } + + +def analyze_sequence( + gaps: np.ndarray, + alphas: list[float], + n_trials: int, + n_baseline: int, + z_min: float, + rng: np.random.Generator, +) -> dict: + original = compute_canonical(gaps) + baseline_vals = {obs: [] for obs in OBS_NAMES} + for _ in range(n_baseline): + obs = compute_canonical(rng.permutation(gaps)) + for name in OBS_NAMES: + baseline_vals[name].append(obs[name]) + + baseline = {} + z = {} + denom = {} + for name in OBS_NAMES: + vals = np.array(baseline_vals[name], dtype=float) + mean = float(np.mean(vals)) + std = float(np.std(vals, ddof=1)) if len(vals) > 1 else 0.0 + baseline[name] = {"mean": mean, "std": std} + denom[name] = float(original[name] - mean) + z[name] = float(denom[name] / std) if std > 1e-15 else 0.0 + + stable_obs = [name for name in OBS_NAMES if abs(z[name]) >= z_min] + all_vectors = [] + screened_vectors = [] + labels = [] + profiles = [] + + for pert_name in PERT_NAMES: + for alpha in alphas: + trial_vals = {obs: [] for obs in OBS_NAMES} + for _ in range(n_trials): + perturbed = PERTURB[pert_name](gaps, alpha, rng) + obs = compute_canonical(perturbed) + for name in OBS_NAMES: + trial_vals[name].append(obs[name]) + means = {name: float(np.mean(trial_vals[name])) for name in OBS_NAMES} + retention = {} + for name in OBS_NAMES: + retention[name] = ( + float((means[name] - baseline[name]["mean"]) / denom[name]) + if abs(denom[name]) > 1e-12 + else 0.0 + ) + all_vector = [retention[name] for name in OBS_NAMES] + screened_vector = [retention[name] for name in stable_obs] + all_vectors.append(all_vector) + if len(stable_obs) >= 2: + screened_vectors.append(screened_vector) + labels.append(pert_name) + profiles.append( + { + "perturbation": pert_name, + "alpha": float(alpha), + "retention": retention, + "retention_vector": all_vector, + } + ) + + all_pca = pca_summary(all_vectors, labels) + screened_pca = pca_summary(screened_vectors, labels) if len(stable_obs) >= 2 else None + + return { + "n_gaps": int(len(gaps)), + "original": original, + "full_shuffle_baseline": baseline, + "denominator": denom, + "original_vs_shuffle_z": z, + "stable_observables": stable_obs, + "weak_observable_count": int(len(OBS_NAMES) - len(stable_obs)), + "profiles": profiles, + "pca_all_observables": all_pca, + "pca_stable_observables": screened_pca, + } + + +def summarize_replicates(items: list[dict]) -> dict: + def arr(path: tuple[str, ...]) -> np.ndarray: + vals = [] + for item in items: + x = item + for key in path: + if x is None: + break + x = x.get(key) + if isinstance(x, (int, float)): + vals.append(float(x)) + return np.array(vals, dtype=float) + + rank = arr(("pca_all_observables", "effective_rank")) + pc2 = arr(("pca_all_observables", "pc2")) + weak = np.array([item["weak_observable_count"] for item in items], dtype=float) + stable_rank = arr(("pca_stable_observables", "effective_rank")) + cos = arr(("pca_all_observables", "centroid_cosine", "adjacent_swap_vs_large_gap_only")) + + out = { + "n_replicates": len(items), + "rank_mean": float(np.mean(rank)) if len(rank) else 0.0, + "rank_std": float(np.std(rank, ddof=1)) if len(rank) > 1 else 0.0, + "pc2_mean": float(np.mean(pc2)) if len(pc2) else 0.0, + "pc2_std": float(np.std(pc2, ddof=1)) if len(pc2) > 1 else 0.0, + "weak_observable_count_mean": float(np.mean(weak)) if len(weak) else 0.0, + "stable_rank_mean": float(np.mean(stable_rank)) if len(stable_rank) else None, + "stable_rank_std": float(np.std(stable_rank, ddof=1)) if len(stable_rank) > 1 else 0.0, + "adjacent_vs_large_cosine_mean": float(np.mean(cos)) if len(cos) else 0.0, + "adjacent_vs_large_cosine_std": float(np.std(cos, ddof=1)) if len(cos) > 1 else 0.0, + } + if len(rank) > 1 and np.std(weak) > 1e-15 and np.std(rank) > 1e-15: + out["rank_vs_weak_count_corr"] = float(np.corrcoef(rank, weak)[0, 1]) + else: + out["rank_vs_weak_count_corr"] = 0.0 + return out + + +def build_prime_windows(max_n: int, n_reps: int) -> list[np.ndarray]: + total = max_n * n_reps + max_n + gaps = prime_gaps(total) + max_start = len(gaps) - max_n + starts = np.linspace(0, max_start, n_reps, dtype=int) + return [gaps[start : start + max_n].astype(float) for start in starts] + + +def run(args: argparse.Namespace) -> dict: + root_rng = np.random.default_rng(args.seed) + sizes = [int(x) for x in args.sizes.split(",") if x.strip()] + max_n = max(sizes) + alphas = [float(x) for x in np.linspace(args.alpha_min, args.alpha_max, args.n_alpha)] + + output = { + "experiment": "perturbation_rank_size_curve", + "question": "Does perturbation effective-rank stabilize with sample size under canonical observables?", + "observables_registry": OBSERVABLES_REGISTRY_VERSION, + "observables_used": OBS_NAMES, + "perturbations": PERT_NAMES, + "params": vars(args), + "alphas": alphas, + "domains": {}, + "summary": {}, + } + + prime_windows = build_prime_windows(max_n, args.n_replicates) + + print(f"observables_registry={OBSERVABLES_REGISTRY_VERSION}") + print(f"{'domain':<22} {'N':>6} {'rank':>7} {'PC2':>7} {'weak':>5} {'stable_rank':>11}") + + domain_builders = { + "primes_windows": lambda rep_rng, rep_i: prime_windows[rep_i], + "prime_shuffle_control": lambda rep_rng, rep_i: rep_rng.permutation(prime_windows[rep_i]), + "poisson": lambda rep_rng, rep_i: rep_rng.exponential(1.0, size=max_n), + "gue": lambda rep_rng, rep_i: gue_spacings(args.gue_matrix_size, max_n, rep_rng), + } + + for domain_name, builder in domain_builders.items(): + output["domains"][domain_name] = {} + output["summary"][domain_name] = {} + bases = [] + for rep_i in range(args.n_replicates): + rep_rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) + bases.append(builder(rep_rng, rep_i)) + + for n in sizes: + rows = [] + for rep_i, base in enumerate(bases): + rep_rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) + res = analyze_sequence( + base[:n], + alphas=alphas, + n_trials=args.n_trials, + n_baseline=args.n_baseline, + z_min=args.z_min, + rng=rep_rng, + ) + res["replicate"] = rep_i + rows.append(res) + summary = summarize_replicates(rows) + output["domains"][domain_name][str(n)] = rows + output["summary"][domain_name][str(n)] = summary + stable_rank = summary["stable_rank_mean"] + stable_text = f"{stable_rank:>11.3f}" if stable_rank is not None else f"{'NA':>11}" + print( + f"{domain_name:<22} {n:>6} {summary['rank_mean']:>7.3f} " + f"{summary['pc2_mean']:>7.3f} {summary['weak_observable_count_mean']:>5.2f} " + f"{stable_text}" + ) + + out_path = Path(args.out) + out_path.parent.mkdir(parents=True, exist_ok=True) + with out_path.open("w") as f: + json.dump(output, f, indent=2) + print(f"saved {out_path}") + return output + + +def main() -> None: + parser = argparse.ArgumentParser() + parser.add_argument("--sizes", default="128,256,512,1024,2048") + parser.add_argument("--n-replicates", type=int, default=8) + parser.add_argument("--gue-matrix-size", type=int, default=180) + parser.add_argument("--n-alpha", type=int, default=5) + parser.add_argument("--alpha-min", type=float, default=0.1) + parser.add_argument("--alpha-max", type=float, default=0.9) + parser.add_argument("--n-trials", type=int, default=8) + parser.add_argument("--n-baseline", type=int, default=16) + parser.add_argument("--z-min", type=float, default=2.0) + parser.add_argument("--seed", type=int, default=20260506) + parser.add_argument("--out", default="tools/data/perturbation_rank_size_curve.json") + args = parser.parse_args() + run(args) + + +if __name__ == "__main__": + main() codex Scrivo il report e cristallizzo nel seme un vincolo stretto: la dimensionalità di perturbazione va riportata insieme alla qualità dei denominatori degli osservabili, altrimenti Poisson e shuffle-primi producono falsi assi comparabili. apply patch patch: completed /opt/MM_D-ND/tools/data/reports/agent_20260506_1941.md diff --git a/tools/data/reports/agent_20260506_1941.md b/tools/data/reports/agent_20260506_1941.md new file mode 100644 index 0000000000000000000000000000000000000000..5d74b086d2d0134f6b5a9dc93f036fd8b669c104 --- /dev/null +++ b/tools/data/reports/agent_20260506_1941.md @@ -0,0 +1,90 @@ +# Agent Report — Perturbation Rank Needs Denominator Gating +**Date**: 2026-05-06 19:41 +**Piano**: 66 +**Tension explored**: META + BOUNDARY +observables_registry: 1.0.0-2026-05-06 +observables_used: [SR, SR2, L1, L2, triple_var] + +## Claim Under Test +The valid next test from cycle 06:25 was a replicate-and-size curve for perturbation effective rank, with observable definitions versioned. This run asks: + +> Does the second perturbation axis stabilize as sample size grows, or is rank inflated when retention denominators are weak relative to full-shuffle baselines? + +## Experiment +Tool created: `tools/exp_perturbation_rank_size_curve.py` + +Atomic perimeter: +- domains: prime-gap windows, prime-shuffle controls, iid Poisson spacings, independent GUE spacings; +- sample sizes: 128, 256, 512, 1024, 2048 gaps; +- replicates/windows: 8 per domain-size point; +- perturbations: `adjacent_swap`, `block_shuffle`, `large_gap_only`, `uniform`; +- alpha grid: 0.1, 0.3, 0.5, 0.7, 0.9; +- trials per perturbation-alpha: 8; +- full-shuffle baselines: 16; +- canonical observables imported from `tools/observables_registry.py`; +- denominator gate: observable is stable only when `abs(original - shuffle_mean) / shuffle_std >= 2`. + +The script reports two ranks: +- `rank_all`: PCA effective rank using all five canonical observables; +- `stable_rank`: PCA effective rank after dropping observables whose original-vs-shuffle denominator is weak. + +## Results + +### Size Curve Summary + +| Domain | N | rank_all | PC2 | weak obs / 5 | stable_rank | +|---|---:|---:|---:|---:|---:| +| primes_windows | 128 | 1.789 ± 0.469 | 0.155 | 4.50 | 1.382 | +| primes_windows | 256 | 1.947 ± 0.645 | 0.174 | 4.75 | 1.262 | +| primes_windows | 512 | 1.892 ± 0.372 | 0.142 | 2.88 | 1.310 | +| primes_windows | 1024 | 1.679 ± 0.409 | 0.117 | 1.62 | 1.415 | +| primes_windows | 2048 | 1.442 ± 0.213 | 0.081 | 0.75 | 1.462 | +| prime_shuffle_control | 2048 | 1.797 ± 0.375 | 0.134 | 3.62 | 1.428 | +| poisson | 2048 | 1.952 ± 0.499 | 0.175 | 4.62 | 1.036 | +| gue | 128 | 1.703 ± 0.348 | 0.126 | 2.38 | 1.226 | +| gue | 256 | 1.913 ± 0.453 | 0.164 | 2.25 | 1.141 | +| gue | 512 | 1.542 ± 0.313 | 0.111 | 1.88 | 1.162 | +| gue | 1024 | 1.551 ± 0.395 | 0.105 | 1.88 | 1.157 | +| gue | 2048 | 1.234 ± 0.224 | 0.046 | 2.00 | 1.111 | + +### Observable Stability + +At GUE N=2048, `SR`, `L1`, and `triple_var` are stable in all 8 replicates; `SR2` and `L2` are stable in 0 of 8. Mean absolute z-scores: `SR=8.38`, `SR2=0.67`, `L1=11.58`, `L2=0.89`, `triple_var=11.66`. + +At primes N=2048, `SR`, `L1`, and `triple_var` are stable in all 8 windows; `SR2` is stable in 7 of 8; `L2` is stable in 3 of 8. Mean absolute z-scores: `SR=5.19`, `SR2=2.63`, `L1=3.96`, `L2=1.78`, `triple_var=4.37`. + +Poisson and prime-shuffle controls keep high `rank_all` while most observables are weak. At Poisson N=2048, `rank_all=1.952` but `stable_rank=1.036` and 4.62 of 5 observables are weak on average. This is the falsifying control for treating rank_all alone as a structural claim. + +## Findings + +1. **Perturbation rank is not interpretable without denominator gating.** In this perimeter, Poisson and prime-shuffle controls can show `rank_all` near 1.8-2.0. Because their original-vs-shuffle denominators are mostly weak, that rank is a retention-normalization artifact unless the stable-observable screen also supports it. + +2. **GUE does not show a stable second axis on canonical observables up to N=2048.** GUE `rank_all` falls from 1.913 at N=256 to 1.234 at N=2048; PC2 falls from 16.4% to 4.6%. After denominator gating, GUE stable rank stays close to 1.1-1.2. + +3. **The old L2-driven sign-flip should not be promoted without a denominator check.** Under canonical observables, GUE `L2` is weak relative to shuffle at every tested size and is stable in 0/8 replicates at N >= 512. This does not prove every L2 sign effect is false; it restricts such effects to local/sample-specific observations unless the denominator survives. + +4. **Primes become better conditioned with N, but not more multi-axis.** Prime windows gain stable observables as N grows: weak count drops from 4.75 at N=256 to 0.75 at N=2048. The effective rank does not grow with this conditioning; it is 1.442 at N=2048, and stable_rank is 1.462. + +## Verdict +**CONSTRAINT on META + BOUNDARY**: perturbation dimensionality must be reported as: + +> effective rank + PC2 + observable registry version + original-vs-shuffle z gate per observable. + +The cycle 03:30 "second GUE axis" remains restricted by cycle 06:25 and is further narrowed here: under canonical observables and the tested size curve, the stable statement is not "GUE has a second perturbation axis"; it is: + +> all-observable perturbation rank can inflate in weak-denominator regimes; after denominator gating, GUE and primes are both close to one perturbation coordinate in this perimeter, while Poisson/shuffle controls show why ungated rank is not structural evidence. + +## Consecutio +What opens now: the lab can keep using perturbation rank, but only as a gated observable. The next useful movement is not more PCA; it is an operator-level denominator map: for each observable, identify the perturbation/domain/scale region where `original - shuffle` is a real signal rather than a noisy divisor. + +## Auto-audit: 5 lenti +- **L1 hard constraint vs bias**: no zero/always claim. "Weak" means `abs(z) < 2` in the declared gate, not absence of signal. +- **L2 quantity vs ratio**: retention ratios are not read alone; raw denominator z-scores are reported first. +- **L3 no silent patching**: the 03:30 claim is explicitly restricted; it is not renamed as confirmed. +- **L4 edge cases**: short-GUE and low-N effects are isolated by size. The N=2048 perimeter is stated, not generalized. +- **L5 re-discovery**: PCA rank inflation from noisy normalization is a standard statistical risk. This is a lab constraint on method, not a new RMT result. + +## Files +- Script: `tools/exp_perturbation_rank_size_curve.py` +- Data: `tools/data/perturbation_rank_size_curve.json` +- Report: `tools/data/reports/agent_20260506_1941.md` diff --git a/tools/exp_perturbation_rank_size_curve.py b/tools/exp_perturbation_rank_size_curve.py new file mode 100644 index 0000000000000000000000000000000000000000..032847ca3d82ec3233ec2b41ad230dfe394d2791 --- /dev/null +++ b/tools/exp_perturbation_rank_size_curve.py @@ -0,0 +1,387 @@ +#!/usr/bin/env python3 +""" +exp_perturbation_rank_size_curve.py + +Reusable META audit for perturbation dimensionality. + +The 2026-05-06 06:25 cycle restricted the claim "GUE has a second +perturbation axis" to sample size, generator, and observable definitions. +This tool measures the size curve directly, using the canonical observable +registry and explicit original-vs-shuffle denominator diagnostics. + +The report owns interpretation. This script only measures: +- effective rank and PC2 across scale-selective perturbation profiles; +- original-vs-shuffle z-score per observable; +- whether apparent rank co-occurs with weak retention denominators. +""" + +from __future__ import annotations + +import argparse +import json +from pathlib import Path + +import numpy as np + +from observables_registry import ( + OBSERVABLES_CANONICAL, + OBSERVABLES_REGISTRY_VERSION, + compute_canonical, +) + + +OBS_NAMES = list(OBSERVABLES_CANONICAL.keys()) +PERT_NAMES = ["adjacent_swap", "block_shuffle", "large_gap_only", "uniform"] + + +def prime_gaps(n_gaps: int) -> np.ndarray: + """Return the first n_gaps prime gaps using a compact numpy sieve.""" + limit = max(100, int(n_gaps * (np.log(n_gaps + 10) + np.log(np.log(n_gaps + 10)) + 5))) + while True: + sieve = np.ones(limit + 1, dtype=bool) + sieve[:2] = False + for p in range(2, int(limit**0.5) + 1): + if sieve[p]: + sieve[p * p : limit + 1 : p] = False + primes = np.flatnonzero(sieve) + if len(primes) >= n_gaps + 1: + return np.diff(primes[: n_gaps + 1]).astype(float) + limit *= 2 + + +def gue_spacings(matrix_size: int, min_spacings: int, rng: np.random.Generator) -> np.ndarray: + """Generate unfolded GUE spacings by concatenating independent matrices.""" + parts = [] + edge = max(2, matrix_size // 10) + while sum(len(x) for x in parts) < min_spacings: + real = rng.standard_normal((matrix_size, matrix_size)) + imag = rng.standard_normal((matrix_size, matrix_size)) + h = real + 1j * imag + h = (h + h.conj().T) / (2.0 * np.sqrt(matrix_size)) + eigs = np.sort(np.linalg.eigvalsh(h).real) + bulk = eigs[edge:-edge] + gaps = np.diff(bulk) + mean = float(np.mean(gaps)) + if mean > 1e-15: + parts.append(gaps / mean) + return np.concatenate(parts)[:min_spacings].astype(float) + + +def perturb_adjacent_swap(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + idx = np.arange(0, len(out) - 1, 2) + chosen = idx[rng.random(len(idx)) < alpha] + tmp = out[chosen].copy() + out[chosen] = out[chosen + 1] + out[chosen + 1] = tmp + return out + + +def perturb_block_shuffle(gaps: np.ndarray, alpha: float, rng: np.random.Generator, block_size: int = 64) -> np.ndarray: + out = gaps.copy() + n_blocks = len(out) // block_size + if n_blocks <= 0: + return rng.permutation(out) if alpha > 0 else out + k = int(round(alpha * n_blocks)) + if k <= 0: + return out + for block in rng.choice(n_blocks, size=min(k, n_blocks), replace=False): + start = block * block_size + end = min(start + block_size, len(out)) + rng.shuffle(out[start:end]) + return out + + +def perturb_large_gap_only(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + idx = np.flatnonzero(out > np.median(out)) + k = int(round(alpha * len(idx))) + if k < 2: + return out + chosen = rng.choice(idx, size=min(k, len(idx)), replace=False) + vals = out[chosen].copy() + rng.shuffle(vals) + out[chosen] = vals + return out + + +def perturb_uniform(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + k = int(round(alpha * len(out))) + if k < 2: + return out + chosen = rng.choice(len(out), size=min(k, len(out)), replace=False) + vals = out[chosen].copy() + rng.shuffle(vals) + out[chosen] = vals + return out + + +PERTURB = { + "adjacent_swap": perturb_adjacent_swap, + "block_shuffle": perturb_block_shuffle, + "large_gap_only": perturb_large_gap_only, + "uniform": perturb_uniform, +} + + +def pca_summary(vectors: list[list[float]], labels: list[str]) -> dict: + matrix = np.array(vectors, dtype=float) + matrix = matrix - np.mean(matrix, axis=0, keepdims=True) + if min(matrix.shape) == 0 or np.max(np.abs(matrix)) <= 1e-15: + return { + "explained_variance": [], + "effective_rank": 0.0, + "centroid_cosine": {}, + "pc2": 0.0, + } + _, singular, _ = np.linalg.svd(matrix, full_matrices=False) + energy = singular * singular + if float(np.sum(energy)) <= 1e-15: + explained = np.zeros_like(energy) + effective_rank = 0.0 + else: + explained = energy / np.sum(energy) + pos = explained[explained > 1e-15] + effective_rank = float(np.exp(-np.sum(pos * np.log(pos)))) + + centroids = {} + for name in PERT_NAMES: + vals = np.array([v for v, label in zip(vectors, labels) if label == name], dtype=float) + if len(vals): + centroids[name] = np.mean(vals, axis=0) + + cosine = {} + for i, a_name in enumerate(PERT_NAMES): + for b_name in PERT_NAMES[i + 1 :]: + if a_name not in centroids or b_name not in centroids: + continue + a = centroids[a_name] + b = centroids[b_name] + denom = np.linalg.norm(a) * np.linalg.norm(b) + cosine[f"{a_name}_vs_{b_name}"] = float(np.dot(a, b) / denom) if denom > 1e-15 else 0.0 + + return { + "explained_variance": [float(x) for x in explained], + "effective_rank": effective_rank, + "centroid_cosine": cosine, + "pc2": float(explained[1]) if len(explained) > 1 else 0.0, + } + + +def analyze_sequence( + gaps: np.ndarray, + alphas: list[float], + n_trials: int, + n_baseline: int, + z_min: float, + rng: np.random.Generator, +) -> dict: + original = compute_canonical(gaps) + baseline_vals = {obs: [] for obs in OBS_NAMES} + for _ in range(n_baseline): + obs = compute_canonical(rng.permutation(gaps)) + for name in OBS_NAMES: + baseline_vals[name].append(obs[name]) + + baseline = {} + z = {} + denom = {} + for name in OBS_NAMES: + vals = np.array(baseline_vals[name], dtype=float) + mean = float(np.mean(vals)) + std = float(np.std(vals, ddof=1)) if len(vals) > 1 else 0.0 + baseline[name] = {"mean": mean, "std": std} + denom[name] = float(original[name] - mean) + z[name] = float(denom[name] / std) if std > 1e-15 else 0.0 + + stable_obs = [name for name in OBS_NAMES if abs(z[name]) >= z_min] + all_vectors = [] + screened_vectors = [] + labels = [] + profiles = [] + + for pert_name in PERT_NAMES: + for alpha in alphas: + trial_vals = {obs: [] for obs in OBS_NAMES} + for _ in range(n_trials): + perturbed = PERTURB[pert_name](gaps, alpha, rng) + obs = compute_canonical(perturbed) + for name in OBS_NAMES: + trial_vals[name].append(obs[name]) + means = {name: float(np.mean(trial_vals[name])) for name in OBS_NAMES} + retention = {} + for name in OBS_NAMES: + retention[name] = ( + float((means[name] - baseline[name]["mean"]) / denom[name]) + if abs(denom[name]) > 1e-12 + else 0.0 + ) + all_vector = [retention[name] for name in OBS_NAMES] + screened_vector = [retention[name] for name in stable_obs] + all_vectors.append(all_vector) + if len(stable_obs) >= 2: + screened_vectors.append(screened_vector) + labels.append(pert_name) + profiles.append( + { + "perturbation": pert_name, + "alpha": float(alpha), + "retention": retention, + "retention_vector": all_vector, + } + ) + + all_pca = pca_summary(all_vectors, labels) + screened_pca = pca_summary(screened_vectors, labels) if len(stable_obs) >= 2 else None + + return { + "n_gaps": int(len(gaps)), + "original": original, + "full_shuffle_baseline": baseline, + "denominator": denom, + "original_vs_shuffle_z": z, + "stable_observables": stable_obs, + "weak_observable_count": int(len(OBS_NAMES) - len(stable_obs)), + "profiles": profiles, + "pca_all_observables": all_pca, + "pca_stable_observables": screened_pca, + } + + +def summarize_replicates(items: list[dict]) -> dict: + def arr(path: tuple[str, ...]) -> np.ndarray: + vals = [] + for item in items: + x = item + for key in path: + if x is None: + break + x = x.get(key) + if isinstance(x, (int, float)): + vals.append(float(x)) + return np.array(vals, dtype=float) + + rank = arr(("pca_all_observables", "effective_rank")) + pc2 = arr(("pca_all_observables", "pc2")) + weak = np.array([item["weak_observable_count"] for item in items], dtype=float) + stable_rank = arr(("pca_stable_observables", "effective_rank")) + cos = arr(("pca_all_observables", "centroid_cosine", "adjacent_swap_vs_large_gap_only")) + + out = { + "n_replicates": len(items), + "rank_mean": float(np.mean(rank)) if len(rank) else 0.0, + "rank_std": float(np.std(rank, ddof=1)) if len(rank) > 1 else 0.0, + "pc2_mean": float(np.mean(pc2)) if len(pc2) else 0.0, + "pc2_std": float(np.std(pc2, ddof=1)) if len(pc2) > 1 else 0.0, + "weak_observable_count_mean": float(np.mean(weak)) if len(weak) else 0.0, + "stable_rank_mean": float(np.mean(stable_rank)) if len(stable_rank) else None, + "stable_rank_std": float(np.std(stable_rank, ddof=1)) if len(stable_rank) > 1 else 0.0, + "adjacent_vs_large_cosine_mean": float(np.mean(cos)) if len(cos) else 0.0, + "adjacent_vs_large_cosine_std": float(np.std(cos, ddof=1)) if len(cos) > 1 else 0.0, + } + if len(rank) > 1 and np.std(weak) > 1e-15 and np.std(rank) > 1e-15: + out["rank_vs_weak_count_corr"] = float(np.corrcoef(rank, weak)[0, 1]) + else: + out["rank_vs_weak_count_corr"] = 0.0 + return out + + +def build_prime_windows(max_n: int, n_reps: int) -> list[np.ndarray]: + total = max_n * n_reps + max_n + gaps = prime_gaps(total) + max_start = len(gaps) - max_n + starts = np.linspace(0, max_start, n_reps, dtype=int) + return [gaps[start : start + max_n].astype(float) for start in starts] + + +def run(args: argparse.Namespace) -> dict: + root_rng = np.random.default_rng(args.seed) + sizes = [int(x) for x in args.sizes.split(",") if x.strip()] + max_n = max(sizes) + alphas = [float(x) for x in np.linspace(args.alpha_min, args.alpha_max, args.n_alpha)] + + output = { + "experiment": "perturbation_rank_size_curve", + "question": "Does perturbation effective-rank stabilize with sample size under canonical observables?", + "observables_registry": OBSERVABLES_REGISTRY_VERSION, + "observables_used": OBS_NAMES, + "perturbations": PERT_NAMES, + "params": vars(args), + "alphas": alphas, + "domains": {}, + "summary": {}, + } + + prime_windows = build_prime_windows(max_n, args.n_replicates) + + print(f"observables_registry={OBSERVABLES_REGISTRY_VERSION}") + print(f"{'domain':<22} {'N':>6} {'rank':>7} {'PC2':>7} {'weak':>5} {'stable_rank':>11}") + + domain_builders = { + "primes_windows": lambda rep_rng, rep_i: prime_windows[rep_i], + "prime_shuffle_control": lambda rep_rng, rep_i: rep_rng.permutation(prime_windows[rep_i]), + "poisson": lambda rep_rng, rep_i: rep_rng.exponential(1.0, size=max_n), + "gue": lambda rep_rng, rep_i: gue_spacings(args.gue_matrix_size, max_n, rep_rng), + } + + for domain_name, builder in domain_builders.items(): + output["domains"][domain_name] = {} + output["summary"][domain_name] = {} + bases = [] + for rep_i in range(args.n_replicates): + rep_rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) + bases.append(builder(rep_rng, rep_i)) + + for n in sizes: + rows = [] + for rep_i, base in enumerate(bases): + rep_rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) + res = analyze_sequence( + base[:n], + alphas=alphas, + n_trials=args.n_trials, + n_baseline=args.n_baseline, + z_min=args.z_min, + rng=rep_rng, + ) + res["replicate"] = rep_i + rows.append(res) + summary = summarize_replicates(rows) + output["domains"][domain_name][str(n)] = rows + output["summary"][domain_name][str(n)] = summary + stable_rank = summary["stable_rank_mean"] + stable_text = f"{stable_rank:>11.3f}" if stable_rank is not None else f"{'NA':>11}" + print( + f"{domain_name:<22} {n:>6} {summary['rank_mean']:>7.3f} " + f"{summary['pc2_mean']:>7.3f} {summary['weak_observable_count_mean']:>5.2f} " + f"{stable_text}" + ) + + out_path = Path(args.out) + out_path.parent.mkdir(parents=True, exist_ok=True) + with out_path.open("w") as f: + json.dump(output, f, indent=2) + print(f"saved {out_path}") + return output + + +def main() -> None: + parser = argparse.ArgumentParser() + parser.add_argument("--sizes", default="128,256,512,1024,2048") + parser.add_argument("--n-replicates", type=int, default=8) + parser.add_argument("--gue-matrix-size", type=int, default=180) + parser.add_argument("--n-alpha", type=int, default=5) + parser.add_argument("--alpha-min", type=float, default=0.1) + parser.add_argument("--alpha-max", type=float, default=0.9) + parser.add_argument("--n-trials", type=int, default=8) + parser.add_argument("--n-baseline", type=int, default=16) + parser.add_argument("--z-min", type=float, default=2.0) + parser.add_argument("--seed", type=int, default=20260506) + parser.add_argument("--out", default="tools/data/perturbation_rank_size_curve.json") + args = parser.parse_args() + run(args) + + +if __name__ == "__main__": + main() diff --git a/tools/data/reports/agent_20260506_1941.md b/tools/data/reports/agent_20260506_1941.md new file mode 100644 index 0000000000000000000000000000000000000000..5d74b086d2d0134f6b5a9dc93f036fd8b669c104 --- /dev/null +++ b/tools/data/reports/agent_20260506_1941.md @@ -0,0 +1,90 @@ +# Agent Report — Perturbation Rank Needs Denominator Gating +**Date**: 2026-05-06 19:41 +**Piano**: 66 +**Tension explored**: META + BOUNDARY +observables_registry: 1.0.0-2026-05-06 +observables_used: [SR, SR2, L1, L2, triple_var] + +## Claim Under Test +The valid next test from cycle 06:25 was a replicate-and-size curve for perturbation effective rank, with observable definitions versioned. This run asks: + +> Does the second perturbation axis stabilize as sample size grows, or is rank inflated when retention denominators are weak relative to full-shuffle baselines? + +## Experiment +Tool created: `tools/exp_perturbation_rank_size_curve.py` + +Atomic perimeter: +- domains: prime-gap windows, prime-shuffle controls, iid Poisson spacings, independent GUE spacings; +- sample sizes: 128, 256, 512, 1024, 2048 gaps; +- replicates/windows: 8 per domain-size point; +- perturbations: `adjacent_swap`, `block_shuffle`, `large_gap_only`, `uniform`; +- alpha grid: 0.1, 0.3, 0.5, 0.7, 0.9; +- trials per perturbation-alpha: 8; +- full-shuffle baselines: 16; +- canonical observables imported from `tools/observables_registry.py`; +- denominator gate: observable is stable only when `abs(original - shuffle_mean) / shuffle_std >= 2`. + +The script reports two ranks: +- `rank_all`: PCA effective rank using all five canonical observables; +- `stable_rank`: PCA effective rank after dropping observables whose original-vs-shuffle denominator is weak. + +## Results + +### Size Curve Summary + +| Domain | N | rank_all | PC2 | weak obs / 5 | stable_rank | +|---|---:|---:|---:|---:|---:| +| primes_windows | 128 | 1.789 ± 0.469 | 0.155 | 4.50 | 1.382 | +| primes_windows | 256 | 1.947 ± 0.645 | 0.174 | 4.75 | 1.262 | +| primes_windows | 512 | 1.892 ± 0.372 | 0.142 | 2.88 | 1.310 | +| primes_windows | 1024 | 1.679 ± 0.409 | 0.117 | 1.62 | 1.415 | +| primes_windows | 2048 | 1.442 ± 0.213 | 0.081 | 0.75 | 1.462 | +| prime_shuffle_control | 2048 | 1.797 ± 0.375 | 0.134 | 3.62 | 1.428 | +| poisson | 2048 | 1.952 ± 0.499 | 0.175 | 4.62 | 1.036 | +| gue | 128 | 1.703 ± 0.348 | 0.126 | 2.38 | 1.226 | +| gue | 256 | 1.913 ± 0.453 | 0.164 | 2.25 | 1.141 | +| gue | 512 | 1.542 ± 0.313 | 0.111 | 1.88 | 1.162 | +| gue | 1024 | 1.551 ± 0.395 | 0.105 | 1.88 | 1.157 | +| gue | 2048 | 1.234 ± 0.224 | 0.046 | 2.00 | 1.111 | + +### Observable Stability + +At GUE N=2048, `SR`, `L1`, and `triple_var` are stable in all 8 replicates; `SR2` and `L2` are stable in 0 of 8. Mean absolute z-scores: `SR=8.38`, `SR2=0.67`, `L1=11.58`, `L2=0.89`, `triple_var=11.66`. + +At primes N=2048, `SR`, `L1`, and `triple_var` are stable in all 8 windows; `SR2` is stable in 7 of 8; `L2` is stable in 3 of 8. Mean absolute z-scores: `SR=5.19`, `SR2=2.63`, `L1=3.96`, `L2=1.78`, `triple_var=4.37`. + +Poisson and prime-shuffle controls keep high `rank_all` while most observables are weak. At Poisson N=2048, `rank_all=1.952` but `stable_rank=1.036` and 4.62 of 5 observables are weak on average. This is the falsifying control for treating rank_all alone as a structural claim. + +## Findings + +1. **Perturbation rank is not interpretable without denominator gating.** In this perimeter, Poisson and prime-shuffle controls can show `rank_all` near 1.8-2.0. Because their original-vs-shuffle denominators are mostly weak, that rank is a retention-normalization artifact unless the stable-observable screen also supports it. + +2. **GUE does not show a stable second axis on canonical observables up to N=2048.** GUE `rank_all` falls from 1.913 at N=256 to 1.234 at N=2048; PC2 falls from 16.4% to 4.6%. After denominator gating, GUE stable rank stays close to 1.1-1.2. + +3. **The old L2-driven sign-flip should not be promoted without a denominator check.** Under canonical observables, GUE `L2` is weak relative to shuffle at every tested size and is stable in 0/8 replicates at N >= 512. This does not prove every L2 sign effect is false; it restricts such effects to local/sample-specific observations unless the denominator survives. + +4. **Primes become better conditioned with N, but not more multi-axis.** Prime windows gain stable observables as N grows: weak count drops from 4.75 at N=256 to 0.75 at N=2048. The effective rank does not grow with this conditioning; it is 1.442 at N=2048, and stable_rank is 1.462. + +## Verdict +**CONSTRAINT on META + BOUNDARY**: perturbation dimensionality must be reported as: + +> effective rank + PC2 + observable registry version + original-vs-shuffle z gate per observable. + +The cycle 03:30 "second GUE axis" remains restricted by cycle 06:25 and is further narrowed here: under canonical observables and the tested size curve, the stable statement is not "GUE has a second perturbation axis"; it is: + +> all-observable perturbation rank can inflate in weak-denominator regimes; after denominator gating, GUE and primes are both close to one perturbation coordinate in this perimeter, while Poisson/shuffle controls show why ungated rank is not structural evidence. + +## Consecutio +What opens now: the lab can keep using perturbation rank, but only as a gated observable. The next useful movement is not more PCA; it is an operator-level denominator map: for each observable, identify the perturbation/domain/scale region where `original - shuffle` is a real signal rather than a noisy divisor. + +## Auto-audit: 5 lenti +- **L1 hard constraint vs bias**: no zero/always claim. "Weak" means `abs(z) < 2` in the declared gate, not absence of signal. +- **L2 quantity vs ratio**: retention ratios are not read alone; raw denominator z-scores are reported first. +- **L3 no silent patching**: the 03:30 claim is explicitly restricted; it is not renamed as confirmed. +- **L4 edge cases**: short-GUE and low-N effects are isolated by size. The N=2048 perimeter is stated, not generalized. +- **L5 re-discovery**: PCA rank inflation from noisy normalization is a standard statistical risk. This is a lab constraint on method, not a new RMT result. + +## Files +- Script: `tools/exp_perturbation_rank_size_curve.py` +- Data: `tools/data/perturbation_rank_size_curve.json` +- Report: `tools/data/reports/agent_20260506_1941.md` diff --git a/tools/exp_perturbation_rank_size_curve.py b/tools/exp_perturbation_rank_size_curve.py new file mode 100644 index 0000000000000000000000000000000000000000..032847ca3d82ec3233ec2b41ad230dfe394d2791 --- /dev/null +++ b/tools/exp_perturbation_rank_size_curve.py @@ -0,0 +1,387 @@ +#!/usr/bin/env python3 +""" +exp_perturbation_rank_size_curve.py + +Reusable META audit for perturbation dimensionality. + +The 2026-05-06 06:25 cycle restricted the claim "GUE has a second +perturbation axis" to sample size, generator, and observable definitions. +This tool measures the size curve directly, using the canonical observable +registry and explicit original-vs-shuffle denominator diagnostics. + +The report owns interpretation. This script only measures: +- effective rank and PC2 across scale-selective perturbation profiles; +- original-vs-shuffle z-score per observable; +- whether apparent rank co-occurs with weak retention denominators. +""" + +from __future__ import annotations + +import argparse +import json +from pathlib import Path + +import numpy as np + +from observables_registry import ( + OBSERVABLES_CANONICAL, + OBSERVABLES_REGISTRY_VERSION, + compute_canonical, +) + + +OBS_NAMES = list(OBSERVABLES_CANONICAL.keys()) +PERT_NAMES = ["adjacent_swap", "block_shuffle", "large_gap_only", "uniform"] + + +def prime_gaps(n_gaps: int) -> np.ndarray: + """Return the first n_gaps prime gaps using a compact numpy sieve.""" + limit = max(100, int(n_gaps * (np.log(n_gaps + 10) + np.log(np.log(n_gaps + 10)) + 5))) + while True: + sieve = np.ones(limit + 1, dtype=bool) + sieve[:2] = False + for p in range(2, int(limit**0.5) + 1): + if sieve[p]: + sieve[p * p : limit + 1 : p] = False + primes = np.flatnonzero(sieve) + if len(primes) >= n_gaps + 1: + return np.diff(primes[: n_gaps + 1]).astype(float) + limit *= 2 + + +def gue_spacings(matrix_size: int, min_spacings: int, rng: np.random.Generator) -> np.ndarray: + """Generate unfolded GUE spacings by concatenating independent matrices.""" + parts = [] + edge = max(2, matrix_size // 10) + while sum(len(x) for x in parts) < min_spacings: + real = rng.standard_normal((matrix_size, matrix_size)) + imag = rng.standard_normal((matrix_size, matrix_size)) + h = real + 1j * imag + h = (h + h.conj().T) / (2.0 * np.sqrt(matrix_size)) + eigs = np.sort(np.linalg.eigvalsh(h).real) + bulk = eigs[edge:-edge] + gaps = np.diff(bulk) + mean = float(np.mean(gaps)) + if mean > 1e-15: + parts.append(gaps / mean) + return np.concatenate(parts)[:min_spacings].astype(float) + + +def perturb_adjacent_swap(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + idx = np.arange(0, len(out) - 1, 2) + chosen = idx[rng.random(len(idx)) < alpha] + tmp = out[chosen].copy() + out[chosen] = out[chosen + 1] + out[chosen + 1] = tmp + return out + + +def perturb_block_shuffle(gaps: np.ndarray, alpha: float, rng: np.random.Generator, block_size: int = 64) -> np.ndarray: + out = gaps.copy() + n_blocks = len(out) // block_size + if n_blocks <= 0: + return rng.permutation(out) if alpha > 0 else out + k = int(round(alpha * n_blocks)) + if k <= 0: + return out + for block in rng.choice(n_blocks, size=min(k, n_blocks), replace=False): + start = block * block_size + end = min(start + block_size, len(out)) + rng.shuffle(out[start:end]) + return out + + +def perturb_large_gap_only(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + idx = np.flatnonzero(out > np.median(out)) + k = int(round(alpha * len(idx))) + if k < 2: + return out + chosen = rng.choice(idx, size=min(k, len(idx)), replace=False) + vals = out[chosen].copy() + rng.shuffle(vals) + out[chosen] = vals + return out + + +def perturb_uniform(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + k = int(round(alpha * len(out))) + if k < 2: + return out + chosen = rng.choice(len(out), size=min(k, len(out)), replace=False) + vals = out[chosen].copy() + rng.shuffle(vals) + out[chosen] = vals + return out + + +PERTURB = { + "adjacent_swap": perturb_adjacent_swap, + "block_shuffle": perturb_block_shuffle, + "large_gap_only": perturb_large_gap_only, + "uniform": perturb_uniform, +} + + +def pca_summary(vectors: list[list[float]], labels: list[str]) -> dict: + matrix = np.array(vectors, dtype=float) + matrix = matrix - np.mean(matrix, axis=0, keepdims=True) + if min(matrix.shape) == 0 or np.max(np.abs(matrix)) <= 1e-15: + return { + "explained_variance": [], + "effective_rank": 0.0, + "centroid_cosine": {}, + "pc2": 0.0, + } + _, singular, _ = np.linalg.svd(matrix, full_matrices=False) + energy = singular * singular + if float(np.sum(energy)) <= 1e-15: + explained = np.zeros_like(energy) + effective_rank = 0.0 + else: + explained = energy / np.sum(energy) + pos = explained[explained > 1e-15] + effective_rank = float(np.exp(-np.sum(pos * np.log(pos)))) + + centroids = {} + for name in PERT_NAMES: + vals = np.array([v for v, label in zip(vectors, labels) if label == name], dtype=float) + if len(vals): + centroids[name] = np.mean(vals, axis=0) + + cosine = {} + for i, a_name in enumerate(PERT_NAMES): + for b_name in PERT_NAMES[i + 1 :]: + if a_name not in centroids or b_name not in centroids: + continue + a = centroids[a_name] + b = centroids[b_name] + denom = np.linalg.norm(a) * np.linalg.norm(b) + cosine[f"{a_name}_vs_{b_name}"] = float(np.dot(a, b) / denom) if denom > 1e-15 else 0.0 + + return { + "explained_variance": [float(x) for x in explained], + "effective_rank": effective_rank, + "centroid_cosine": cosine, + "pc2": float(explained[1]) if len(explained) > 1 else 0.0, + } + + +def analyze_sequence( + gaps: np.ndarray, + alphas: list[float], + n_trials: int, + n_baseline: int, + z_min: float, + rng: np.random.Generator, +) -> dict: + original = compute_canonical(gaps) + baseline_vals = {obs: [] for obs in OBS_NAMES} + for _ in range(n_baseline): + obs = compute_canonical(rng.permutation(gaps)) + for name in OBS_NAMES: + baseline_vals[name].append(obs[name]) + + baseline = {} + z = {} + denom = {} + for name in OBS_NAMES: + vals = np.array(baseline_vals[name], dtype=float) + mean = float(np.mean(vals)) + std = float(np.std(vals, ddof=1)) if len(vals) > 1 else 0.0 + baseline[name] = {"mean": mean, "std": std} + denom[name] = float(original[name] - mean) + z[name] = float(denom[name] / std) if std > 1e-15 else 0.0 + + stable_obs = [name for name in OBS_NAMES if abs(z[name]) >= z_min] + all_vectors = [] + screened_vectors = [] + labels = [] + profiles = [] + + for pert_name in PERT_NAMES: + for alpha in alphas: + trial_vals = {obs: [] for obs in OBS_NAMES} + for _ in range(n_trials): + perturbed = PERTURB[pert_name](gaps, alpha, rng) + obs = compute_canonical(perturbed) + for name in OBS_NAMES: + trial_vals[name].append(obs[name]) + means = {name: float(np.mean(trial_vals[name])) for name in OBS_NAMES} + retention = {} + for name in OBS_NAMES: + retention[name] = ( + float((means[name] - baseline[name]["mean"]) / denom[name]) + if abs(denom[name]) > 1e-12 + else 0.0 + ) + all_vector = [retention[name] for name in OBS_NAMES] + screened_vector = [retention[name] for name in stable_obs] + all_vectors.append(all_vector) + if len(stable_obs) >= 2: + screened_vectors.append(screened_vector) + labels.append(pert_name) + profiles.append( + { + "perturbation": pert_name, + "alpha": float(alpha), + "retention": retention, + "retention_vector": all_vector, + } + ) + + all_pca = pca_summary(all_vectors, labels) + screened_pca = pca_summary(screened_vectors, labels) if len(stable_obs) >= 2 else None + + return { + "n_gaps": int(len(gaps)), + "original": original, + "full_shuffle_baseline": baseline, + "denominator": denom, + "original_vs_shuffle_z": z, + "stable_observables": stable_obs, + "weak_observable_count": int(len(OBS_NAMES) - len(stable_obs)), + "profiles": profiles, + "pca_all_observables": all_pca, + "pca_stable_observables": screened_pca, + } + + +def summarize_replicates(items: list[dict]) -> dict: + def arr(path: tuple[str, ...]) -> np.ndarray: + vals = [] + for item in items: + x = item + for key in path: + if x is None: + break + x = x.get(key) + if isinstance(x, (int, float)): + vals.append(float(x)) + return np.array(vals, dtype=float) + + rank = arr(("pca_all_observables", "effective_rank")) + pc2 = arr(("pca_all_observables", "pc2")) + weak = np.array([item["weak_observable_count"] for item in items], dtype=float) + stable_rank = arr(("pca_stable_observables", "effective_rank")) + cos = arr(("pca_all_observables", "centroid_cosine", "adjacent_swap_vs_large_gap_only")) + + out = { + "n_replicates": len(items), + "rank_mean": float(np.mean(rank)) if len(rank) else 0.0, + "rank_std": float(np.std(rank, ddof=1)) if len(rank) > 1 else 0.0, + "pc2_mean": float(np.mean(pc2)) if len(pc2) else 0.0, + "pc2_std": float(np.std(pc2, ddof=1)) if len(pc2) > 1 else 0.0, + "weak_observable_count_mean": float(np.mean(weak)) if len(weak) else 0.0, + "stable_rank_mean": float(np.mean(stable_rank)) if len(stable_rank) else None, + "stable_rank_std": float(np.std(stable_rank, ddof=1)) if len(stable_rank) > 1 else 0.0, + "adjacent_vs_large_cosine_mean": float(np.mean(cos)) if len(cos) else 0.0, + "adjacent_vs_large_cosine_std": float(np.std(cos, ddof=1)) if len(cos) > 1 else 0.0, + } + if len(rank) > 1 and np.std(weak) > 1e-15 and np.std(rank) > 1e-15: + out["rank_vs_weak_count_corr"] = float(np.corrcoef(rank, weak)[0, 1]) + else: + out["rank_vs_weak_count_corr"] = 0.0 + return out + + +def build_prime_windows(max_n: int, n_reps: int) -> list[np.ndarray]: + total = max_n * n_reps + max_n + gaps = prime_gaps(total) + max_start = len(gaps) - max_n + starts = np.linspace(0, max_start, n_reps, dtype=int) + return [gaps[start : start + max_n].astype(float) for start in starts] + + +def run(args: argparse.Namespace) -> dict: + root_rng = np.random.default_rng(args.seed) + sizes = [int(x) for x in args.sizes.split(",") if x.strip()] + max_n = max(sizes) + alphas = [float(x) for x in np.linspace(args.alpha_min, args.alpha_max, args.n_alpha)] + + output = { + "experiment": "perturbation_rank_size_curve", + "question": "Does perturbation effective-rank stabilize with sample size under canonical observables?", + "observables_registry": OBSERVABLES_REGISTRY_VERSION, + "observables_used": OBS_NAMES, + "perturbations": PERT_NAMES, + "params": vars(args), + "alphas": alphas, + "domains": {}, + "summary": {}, + } + + prime_windows = build_prime_windows(max_n, args.n_replicates) + + print(f"observables_registry={OBSERVABLES_REGISTRY_VERSION}") + print(f"{'domain':<22} {'N':>6} {'rank':>7} {'PC2':>7} {'weak':>5} {'stable_rank':>11}") + + domain_builders = { + "primes_windows": lambda rep_rng, rep_i: prime_windows[rep_i], + "prime_shuffle_control": lambda rep_rng, rep_i: rep_rng.permutation(prime_windows[rep_i]), + "poisson": lambda rep_rng, rep_i: rep_rng.exponential(1.0, size=max_n), + "gue": lambda rep_rng, rep_i: gue_spacings(args.gue_matrix_size, max_n, rep_rng), + } + + for domain_name, builder in domain_builders.items(): + output["domains"][domain_name] = {} + output["summary"][domain_name] = {} + bases = [] + for rep_i in range(args.n_replicates): + rep_rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) + bases.append(builder(rep_rng, rep_i)) + + for n in sizes: + rows = [] + for rep_i, base in enumerate(bases): + rep_rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) + res = analyze_sequence( + base[:n], + alphas=alphas, + n_trials=args.n_trials, + n_baseline=args.n_baseline, + z_min=args.z_min, + rng=rep_rng, + ) + res["replicate"] = rep_i + rows.append(res) + summary = summarize_replicates(rows) + output["domains"][domain_name][str(n)] = rows + output["summary"][domain_name][str(n)] = summary + stable_rank = summary["stable_rank_mean"] + stable_text = f"{stable_rank:>11.3f}" if stable_rank is not None else f"{'NA':>11}" + print( + f"{domain_name:<22} {n:>6} {summary['rank_mean']:>7.3f} " + f"{summary['pc2_mean']:>7.3f} {summary['weak_observable_count_mean']:>5.2f} " + f"{stable_text}" + ) + + out_path = Path(args.out) + out_path.parent.mkdir(parents=True, exist_ok=True) + with out_path.open("w") as f: + json.dump(output, f, indent=2) + print(f"saved {out_path}") + return output + + +def main() -> None: + parser = argparse.ArgumentParser() + parser.add_argument("--sizes", default="128,256,512,1024,2048") + parser.add_argument("--n-replicates", type=int, default=8) + parser.add_argument("--gue-matrix-size", type=int, default=180) + parser.add_argument("--n-alpha", type=int, default=5) + parser.add_argument("--alpha-min", type=float, default=0.1) + parser.add_argument("--alpha-max", type=float, default=0.9) + parser.add_argument("--n-trials", type=int, default=8) + parser.add_argument("--n-baseline", type=int, default=16) + parser.add_argument("--z-min", type=float, default=2.0) + parser.add_argument("--seed", type=int, default=20260506) + parser.add_argument("--out", default="tools/data/perturbation_rank_size_curve.json") + args = parser.parse_args() + run(args) + + +if __name__ == "__main__": + main() apply patch patch: completed /opt/MM_D-ND/tools/data/seme.json diff --git a/tools/data/reports/agent_20260506_1941.md b/tools/data/reports/agent_20260506_1941.md new file mode 100644 index 0000000000000000000000000000000000000000..5d74b086d2d0134f6b5a9dc93f036fd8b669c104 --- /dev/null +++ b/tools/data/reports/agent_20260506_1941.md @@ -0,0 +1,90 @@ +# Agent Report — Perturbation Rank Needs Denominator Gating +**Date**: 2026-05-06 19:41 +**Piano**: 66 +**Tension explored**: META + BOUNDARY +observables_registry: 1.0.0-2026-05-06 +observables_used: [SR, SR2, L1, L2, triple_var] + +## Claim Under Test +The valid next test from cycle 06:25 was a replicate-and-size curve for perturbation effective rank, with observable definitions versioned. This run asks: + +> Does the second perturbation axis stabilize as sample size grows, or is rank inflated when retention denominators are weak relative to full-shuffle baselines? + +## Experiment +Tool created: `tools/exp_perturbation_rank_size_curve.py` + +Atomic perimeter: +- domains: prime-gap windows, prime-shuffle controls, iid Poisson spacings, independent GUE spacings; +- sample sizes: 128, 256, 512, 1024, 2048 gaps; +- replicates/windows: 8 per domain-size point; +- perturbations: `adjacent_swap`, `block_shuffle`, `large_gap_only`, `uniform`; +- alpha grid: 0.1, 0.3, 0.5, 0.7, 0.9; +- trials per perturbation-alpha: 8; +- full-shuffle baselines: 16; +- canonical observables imported from `tools/observables_registry.py`; +- denominator gate: observable is stable only when `abs(original - shuffle_mean) / shuffle_std >= 2`. + +The script reports two ranks: +- `rank_all`: PCA effective rank using all five canonical observables; +- `stable_rank`: PCA effective rank after dropping observables whose original-vs-shuffle denominator is weak. + +## Results + +### Size Curve Summary + +| Domain | N | rank_all | PC2 | weak obs / 5 | stable_rank | +|---|---:|---:|---:|---:|---:| +| primes_windows | 128 | 1.789 ± 0.469 | 0.155 | 4.50 | 1.382 | +| primes_windows | 256 | 1.947 ± 0.645 | 0.174 | 4.75 | 1.262 | +| primes_windows | 512 | 1.892 ± 0.372 | 0.142 | 2.88 | 1.310 | +| primes_windows | 1024 | 1.679 ± 0.409 | 0.117 | 1.62 | 1.415 | +| primes_windows | 2048 | 1.442 ± 0.213 | 0.081 | 0.75 | 1.462 | +| prime_shuffle_control | 2048 | 1.797 ± 0.375 | 0.134 | 3.62 | 1.428 | +| poisson | 2048 | 1.952 ± 0.499 | 0.175 | 4.62 | 1.036 | +| gue | 128 | 1.703 ± 0.348 | 0.126 | 2.38 | 1.226 | +| gue | 256 | 1.913 ± 0.453 | 0.164 | 2.25 | 1.141 | +| gue | 512 | 1.542 ± 0.313 | 0.111 | 1.88 | 1.162 | +| gue | 1024 | 1.551 ± 0.395 | 0.105 | 1.88 | 1.157 | +| gue | 2048 | 1.234 ± 0.224 | 0.046 | 2.00 | 1.111 | + +### Observable Stability + +At GUE N=2048, `SR`, `L1`, and `triple_var` are stable in all 8 replicates; `SR2` and `L2` are stable in 0 of 8. Mean absolute z-scores: `SR=8.38`, `SR2=0.67`, `L1=11.58`, `L2=0.89`, `triple_var=11.66`. + +At primes N=2048, `SR`, `L1`, and `triple_var` are stable in all 8 windows; `SR2` is stable in 7 of 8; `L2` is stable in 3 of 8. Mean absolute z-scores: `SR=5.19`, `SR2=2.63`, `L1=3.96`, `L2=1.78`, `triple_var=4.37`. + +Poisson and prime-shuffle controls keep high `rank_all` while most observables are weak. At Poisson N=2048, `rank_all=1.952` but `stable_rank=1.036` and 4.62 of 5 observables are weak on average. This is the falsifying control for treating rank_all alone as a structural claim. + +## Findings + +1. **Perturbation rank is not interpretable without denominator gating.** In this perimeter, Poisson and prime-shuffle controls can show `rank_all` near 1.8-2.0. Because their original-vs-shuffle denominators are mostly weak, that rank is a retention-normalization artifact unless the stable-observable screen also supports it. + +2. **GUE does not show a stable second axis on canonical observables up to N=2048.** GUE `rank_all` falls from 1.913 at N=256 to 1.234 at N=2048; PC2 falls from 16.4% to 4.6%. After denominator gating, GUE stable rank stays close to 1.1-1.2. + +3. **The old L2-driven sign-flip should not be promoted without a denominator check.** Under canonical observables, GUE `L2` is weak relative to shuffle at every tested size and is stable in 0/8 replicates at N >= 512. This does not prove every L2 sign effect is false; it restricts such effects to local/sample-specific observations unless the denominator survives. + +4. **Primes become better conditioned with N, but not more multi-axis.** Prime windows gain stable observables as N grows: weak count drops from 4.75 at N=256 to 0.75 at N=2048. The effective rank does not grow with this conditioning; it is 1.442 at N=2048, and stable_rank is 1.462. + +## Verdict +**CONSTRAINT on META + BOUNDARY**: perturbation dimensionality must be reported as: + +> effective rank + PC2 + observable registry version + original-vs-shuffle z gate per observable. + +The cycle 03:30 "second GUE axis" remains restricted by cycle 06:25 and is further narrowed here: under canonical observables and the tested size curve, the stable statement is not "GUE has a second perturbation axis"; it is: + +> all-observable perturbation rank can inflate in weak-denominator regimes; after denominator gating, GUE and primes are both close to one perturbation coordinate in this perimeter, while Poisson/shuffle controls show why ungated rank is not structural evidence. + +## Consecutio +What opens now: the lab can keep using perturbation rank, but only as a gated observable. The next useful movement is not more PCA; it is an operator-level denominator map: for each observable, identify the perturbation/domain/scale region where `original - shuffle` is a real signal rather than a noisy divisor. + +## Auto-audit: 5 lenti +- **L1 hard constraint vs bias**: no zero/always claim. "Weak" means `abs(z) < 2` in the declared gate, not absence of signal. +- **L2 quantity vs ratio**: retention ratios are not read alone; raw denominator z-scores are reported first. +- **L3 no silent patching**: the 03:30 claim is explicitly restricted; it is not renamed as confirmed. +- **L4 edge cases**: short-GUE and low-N effects are isolated by size. The N=2048 perimeter is stated, not generalized. +- **L5 re-discovery**: PCA rank inflation from noisy normalization is a standard statistical risk. This is a lab constraint on method, not a new RMT result. + +## Files +- Script: `tools/exp_perturbation_rank_size_curve.py` +- Data: `tools/data/perturbation_rank_size_curve.json` +- Report: `tools/data/reports/agent_20260506_1941.md` diff --git a/tools/data/seme.json b/tools/data/seme.json index 9a30fb093a0fab963e44d5bed81d349bc1755ce9..d2b2b5f07ec443b64416cb27616b846fcb5403dd --- a/tools/data/seme.json +++ b/tools/data/seme.json @@ -98,6 +98,17 @@ "condensato_ref": "A14,A8", "origine": "cristallizzato 06/05 dalla consecutio del cycle 20260506_0625 (autopoietico self-finding)", "added_at": "2026-05-06T07:03:58.213606+00:00" + }, + { + "id": "PERTURBATION_DENOMINATOR_GATE", + "tipo": "vincolo", + "claim": "La dimensionalita di perturbazione va riportata solo insieme a PC2, versione observables_registry e gate original-vs-shuffle per osservabile. Nel perimetro 20260506_1941, Poisson e shuffle-primi producono rank_all ~1.8-2.0 con denominatori deboli; dopo gate abs(z)>=2 il rank stabile torna vicino a 1. Rank PCA non gated non e evidenza strutturale.", + "intensita": 0.95, + "porta": "META_BOUNDARY", + "manuale": true, + "condensato_ref": "A4,A8,A14,C2", + "origine": "cycle agent_20260506_1941: perturbation rank size curve canonical observables", + "added_at": "2026-05-06T19:41:00+00:00" } ], "potenziale_bloccato": [], @@ -276,4 +287,4 @@ "timestamp": "2026-04-04T19:34:08.014Z" } ] -} \ No newline at end of file +} diff --git a/tools/exp_perturbation_rank_size_curve.py b/tools/exp_perturbation_rank_size_curve.py new file mode 100644 index 0000000000000000000000000000000000000000..032847ca3d82ec3233ec2b41ad230dfe394d2791 --- /dev/null +++ b/tools/exp_perturbation_rank_size_curve.py @@ -0,0 +1,387 @@ +#!/usr/bin/env python3 +""" +exp_perturbation_rank_size_curve.py + +Reusable META audit for perturbation dimensionality. + +The 2026-05-06 06:25 cycle restricted the claim "GUE has a second +perturbation axis" to sample size, generator, and observable definitions. +This tool measures the size curve directly, using the canonical observable +registry and explicit original-vs-shuffle denominator diagnostics. + +The report owns interpretation. This script only measures: +- effective rank and PC2 across scale-selective perturbation profiles; +- original-vs-shuffle z-score per observable; +- whether apparent rank co-occurs with weak retention denominators. +""" + +from __future__ import annotations + +import argparse +import json +from pathlib import Path + +import numpy as np + +from observables_registry import ( + OBSERVABLES_CANONICAL, + OBSERVABLES_REGISTRY_VERSION, + compute_canonical, +) + + +OBS_NAMES = list(OBSERVABLES_CANONICAL.keys()) +PERT_NAMES = ["adjacent_swap", "block_shuffle", "large_gap_only", "uniform"] + + +def prime_gaps(n_gaps: int) -> np.ndarray: + """Return the first n_gaps prime gaps using a compact numpy sieve.""" + limit = max(100, int(n_gaps * (np.log(n_gaps + 10) + np.log(np.log(n_gaps + 10)) + 5))) + while True: + sieve = np.ones(limit + 1, dtype=bool) + sieve[:2] = False + for p in range(2, int(limit**0.5) + 1): + if sieve[p]: + sieve[p * p : limit + 1 : p] = False + primes = np.flatnonzero(sieve) + if len(primes) >= n_gaps + 1: + return np.diff(primes[: n_gaps + 1]).astype(float) + limit *= 2 + + +def gue_spacings(matrix_size: int, min_spacings: int, rng: np.random.Generator) -> np.ndarray: + """Generate unfolded GUE spacings by concatenating independent matrices.""" + parts = [] + edge = max(2, matrix_size // 10) + while sum(len(x) for x in parts) < min_spacings: + real = rng.standard_normal((matrix_size, matrix_size)) + imag = rng.standard_normal((matrix_size, matrix_size)) + h = real + 1j * imag + h = (h + h.conj().T) / (2.0 * np.sqrt(matrix_size)) + eigs = np.sort(np.linalg.eigvalsh(h).real) + bulk = eigs[edge:-edge] + gaps = np.diff(bulk) + mean = float(np.mean(gaps)) + if mean > 1e-15: + parts.append(gaps / mean) + return np.concatenate(parts)[:min_spacings].astype(float) + + +def perturb_adjacent_swap(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + idx = np.arange(0, len(out) - 1, 2) + chosen = idx[rng.random(len(idx)) < alpha] + tmp = out[chosen].copy() + out[chosen] = out[chosen + 1] + out[chosen + 1] = tmp + return out + + +def perturb_block_shuffle(gaps: np.ndarray, alpha: float, rng: np.random.Generator, block_size: int = 64) -> np.ndarray: + out = gaps.copy() + n_blocks = len(out) // block_size + if n_blocks <= 0: + return rng.permutation(out) if alpha > 0 else out + k = int(round(alpha * n_blocks)) + if k <= 0: + return out + for block in rng.choice(n_blocks, size=min(k, n_blocks), replace=False): + start = block * block_size + end = min(start + block_size, len(out)) + rng.shuffle(out[start:end]) + return out + + +def perturb_large_gap_only(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + idx = np.flatnonzero(out > np.median(out)) + k = int(round(alpha * len(idx))) + if k < 2: + return out + chosen = rng.choice(idx, size=min(k, len(idx)), replace=False) + vals = out[chosen].copy() + rng.shuffle(vals) + out[chosen] = vals + return out + + +def perturb_uniform(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + k = int(round(alpha * len(out))) + if k < 2: + return out + chosen = rng.choice(len(out), size=min(k, len(out)), replace=False) + vals = out[chosen].copy() + rng.shuffle(vals) + out[chosen] = vals + return out + + +PERTURB = { + "adjacent_swap": perturb_adjacent_swap, + "block_shuffle": perturb_block_shuffle, + "large_gap_only": perturb_large_gap_only, + "uniform": perturb_uniform, +} + + +def pca_summary(vectors: list[list[float]], labels: list[str]) -> dict: + matrix = np.array(vectors, dtype=float) + matrix = matrix - np.mean(matrix, axis=0, keepdims=True) + if min(matrix.shape) == 0 or np.max(np.abs(matrix)) <= 1e-15: + return { + "explained_variance": [], + "effective_rank": 0.0, + "centroid_cosine": {}, + "pc2": 0.0, + } + _, singular, _ = np.linalg.svd(matrix, full_matrices=False) + energy = singular * singular + if float(np.sum(energy)) <= 1e-15: + explained = np.zeros_like(energy) + effective_rank = 0.0 + else: + explained = energy / np.sum(energy) + pos = explained[explained > 1e-15] + effective_rank = float(np.exp(-np.sum(pos * np.log(pos)))) + + centroids = {} + for name in PERT_NAMES: + vals = np.array([v for v, label in zip(vectors, labels) if label == name], dtype=float) + if len(vals): + centroids[name] = np.mean(vals, axis=0) + + cosine = {} + for i, a_name in enumerate(PERT_NAMES): + for b_name in PERT_NAMES[i + 1 :]: + if a_name not in centroids or b_name not in centroids: + continue + a = centroids[a_name] + b = centroids[b_name] + denom = np.linalg.norm(a) * np.linalg.norm(b) + cosine[f"{a_name}_vs_{b_name}"] = float(np.dot(a, b) / denom) if denom > 1e-15 else 0.0 + + return { + "explained_variance": [float(x) for x in explained], + "effective_rank": effective_rank, + "centroid_cosine": cosine, + "pc2": float(explained[1]) if len(explained) > 1 else 0.0, + } + + +def analyze_sequence( + gaps: np.ndarray, + alphas: list[float], + n_trials: int, + n_baseline: int, + z_min: float, + rng: np.random.Generator, +) -> dict: + original = compute_canonical(gaps) + baseline_vals = {obs: [] for obs in OBS_NAMES} + for _ in range(n_baseline): + obs = compute_canonical(rng.permutation(gaps)) + for name in OBS_NAMES: + baseline_vals[name].append(obs[name]) + + baseline = {} + z = {} + denom = {} + for name in OBS_NAMES: + vals = np.array(baseline_vals[name], dtype=float) + mean = float(np.mean(vals)) + std = float(np.std(vals, ddof=1)) if len(vals) > 1 else 0.0 + baseline[name] = {"mean": mean, "std": std} + denom[name] = float(original[name] - mean) + z[name] = float(denom[name] / std) if std > 1e-15 else 0.0 + + stable_obs = [name for name in OBS_NAMES if abs(z[name]) >= z_min] + all_vectors = [] + screened_vectors = [] + labels = [] + profiles = [] + + for pert_name in PERT_NAMES: + for alpha in alphas: + trial_vals = {obs: [] for obs in OBS_NAMES} + for _ in range(n_trials): + perturbed = PERTURB[pert_name](gaps, alpha, rng) + obs = compute_canonical(perturbed) + for name in OBS_NAMES: + trial_vals[name].append(obs[name]) + means = {name: float(np.mean(trial_vals[name])) for name in OBS_NAMES} + retention = {} + for name in OBS_NAMES: + retention[name] = ( + float((means[name] - baseline[name]["mean"]) / denom[name]) + if abs(denom[name]) > 1e-12 + else 0.0 + ) + all_vector = [retention[name] for name in OBS_NAMES] + screened_vector = [retention[name] for name in stable_obs] + all_vectors.append(all_vector) + if len(stable_obs) >= 2: + screened_vectors.append(screened_vector) + labels.append(pert_name) + profiles.append( + { + "perturbation": pert_name, + "alpha": float(alpha), + "retention": retention, + "retention_vector": all_vector, + } + ) + + all_pca = pca_summary(all_vectors, labels) + screened_pca = pca_summary(screened_vectors, labels) if len(stable_obs) >= 2 else None + + return { + "n_gaps": int(len(gaps)), + "original": original, + "full_shuffle_baseline": baseline, + "denominator": denom, + "original_vs_shuffle_z": z, + "stable_observables": stable_obs, + "weak_observable_count": int(len(OBS_NAMES) - len(stable_obs)), + "profiles": profiles, + "pca_all_observables": all_pca, + "pca_stable_observables": screened_pca, + } + + +def summarize_replicates(items: list[dict]) -> dict: + def arr(path: tuple[str, ...]) -> np.ndarray: + vals = [] + for item in items: + x = item + for key in path: + if x is None: + break + x = x.get(key) + if isinstance(x, (int, float)): + vals.append(float(x)) + return np.array(vals, dtype=float) + + rank = arr(("pca_all_observables", "effective_rank")) + pc2 = arr(("pca_all_observables", "pc2")) + weak = np.array([item["weak_observable_count"] for item in items], dtype=float) + stable_rank = arr(("pca_stable_observables", "effective_rank")) + cos = arr(("pca_all_observables", "centroid_cosine", "adjacent_swap_vs_large_gap_only")) + + out = { + "n_replicates": len(items), + "rank_mean": float(np.mean(rank)) if len(rank) else 0.0, + "rank_std": float(np.std(rank, ddof=1)) if len(rank) > 1 else 0.0, + "pc2_mean": float(np.mean(pc2)) if len(pc2) else 0.0, + "pc2_std": float(np.std(pc2, ddof=1)) if len(pc2) > 1 else 0.0, + "weak_observable_count_mean": float(np.mean(weak)) if len(weak) else 0.0, + "stable_rank_mean": float(np.mean(stable_rank)) if len(stable_rank) else None, + "stable_rank_std": float(np.std(stable_rank, ddof=1)) if len(stable_rank) > 1 else 0.0, + "adjacent_vs_large_cosine_mean": float(np.mean(cos)) if len(cos) else 0.0, + "adjacent_vs_large_cosine_std": float(np.std(cos, ddof=1)) if len(cos) > 1 else 0.0, + } + if len(rank) > 1 and np.std(weak) > 1e-15 and np.std(rank) > 1e-15: + out["rank_vs_weak_count_corr"] = float(np.corrcoef(rank, weak)[0, 1]) + else: + out["rank_vs_weak_count_corr"] = 0.0 + return out + + +def build_prime_windows(max_n: int, n_reps: int) -> list[np.ndarray]: + total = max_n * n_reps + max_n + gaps = prime_gaps(total) + max_start = len(gaps) - max_n + starts = np.linspace(0, max_start, n_reps, dtype=int) + return [gaps[start : start + max_n].astype(float) for start in starts] + + +def run(args: argparse.Namespace) -> dict: + root_rng = np.random.default_rng(args.seed) + sizes = [int(x) for x in args.sizes.split(",") if x.strip()] + max_n = max(sizes) + alphas = [float(x) for x in np.linspace(args.alpha_min, args.alpha_max, args.n_alpha)] + + output = { + "experiment": "perturbation_rank_size_curve", + "question": "Does perturbation effective-rank stabilize with sample size under canonical observables?", + "observables_registry": OBSERVABLES_REGISTRY_VERSION, + "observables_used": OBS_NAMES, + "perturbations": PERT_NAMES, + "params": vars(args), + "alphas": alphas, + "domains": {}, + "summary": {}, + } + + prime_windows = build_prime_windows(max_n, args.n_replicates) + + print(f"observables_registry={OBSERVABLES_REGISTRY_VERSION}") + print(f"{'domain':<22} {'N':>6} {'rank':>7} {'PC2':>7} {'weak':>5} {'stable_rank':>11}") + + domain_builders = { + "primes_windows": lambda rep_rng, rep_i: prime_windows[rep_i], + "prime_shuffle_control": lambda rep_rng, rep_i: rep_rng.permutation(prime_windows[rep_i]), + "poisson": lambda rep_rng, rep_i: rep_rng.exponential(1.0, size=max_n), + "gue": lambda rep_rng, rep_i: gue_spacings(args.gue_matrix_size, max_n, rep_rng), + } + + for domain_name, builder in domain_builders.items(): + output["domains"][domain_name] = {} + output["summary"][domain_name] = {} + bases = [] + for rep_i in range(args.n_replicates): + rep_rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) + bases.append(builder(rep_rng, rep_i)) + + for n in sizes: + rows = [] + for rep_i, base in enumerate(bases): + rep_rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) + res = analyze_sequence( + base[:n], + alphas=alphas, + n_trials=args.n_trials, + n_baseline=args.n_baseline, + z_min=args.z_min, + rng=rep_rng, + ) + res["replicate"] = rep_i + rows.append(res) + summary = summarize_replicates(rows) + output["domains"][domain_name][str(n)] = rows + output["summary"][domain_name][str(n)] = summary + stable_rank = summary["stable_rank_mean"] + stable_text = f"{stable_rank:>11.3f}" if stable_rank is not None else f"{'NA':>11}" + print( + f"{domain_name:<22} {n:>6} {summary['rank_mean']:>7.3f} " + f"{summary['pc2_mean']:>7.3f} {summary['weak_observable_count_mean']:>5.2f} " + f"{stable_text}" + ) + + out_path = Path(args.out) + out_path.parent.mkdir(parents=True, exist_ok=True) + with out_path.open("w") as f: + json.dump(output, f, indent=2) + print(f"saved {out_path}") + return output + + +def main() -> None: + parser = argparse.ArgumentParser() + parser.add_argument("--sizes", default="128,256,512,1024,2048") + parser.add_argument("--n-replicates", type=int, default=8) + parser.add_argument("--gue-matrix-size", type=int, default=180) + parser.add_argument("--n-alpha", type=int, default=5) + parser.add_argument("--alpha-min", type=float, default=0.1) + parser.add_argument("--alpha-max", type=float, default=0.9) + parser.add_argument("--n-trials", type=int, default=8) + parser.add_argument("--n-baseline", type=int, default=16) + parser.add_argument("--z-min", type=float, default=2.0) + parser.add_argument("--seed", type=int, default=20260506) + parser.add_argument("--out", default="tools/data/perturbation_rank_size_curve.json") + args = parser.parse_args() + run(args) + + +if __name__ == "__main__": + main() diff --git a/tools/data/reports/agent_20260506_1941.md b/tools/data/reports/agent_20260506_1941.md new file mode 100644 index 0000000000000000000000000000000000000000..5d74b086d2d0134f6b5a9dc93f036fd8b669c104 --- /dev/null +++ b/tools/data/reports/agent_20260506_1941.md @@ -0,0 +1,90 @@ +# Agent Report — Perturbation Rank Needs Denominator Gating +**Date**: 2026-05-06 19:41 +**Piano**: 66 +**Tension explored**: META + BOUNDARY +observables_registry: 1.0.0-2026-05-06 +observables_used: [SR, SR2, L1, L2, triple_var] + +## Claim Under Test +The valid next test from cycle 06:25 was a replicate-and-size curve for perturbation effective rank, with observable definitions versioned. This run asks: + +> Does the second perturbation axis stabilize as sample size grows, or is rank inflated when retention denominators are weak relative to full-shuffle baselines? + +## Experiment +Tool created: `tools/exp_perturbation_rank_size_curve.py` + +Atomic perimeter: +- domains: prime-gap windows, prime-shuffle controls, iid Poisson spacings, independent GUE spacings; +- sample sizes: 128, 256, 512, 1024, 2048 gaps; +- replicates/windows: 8 per domain-size point; +- perturbations: `adjacent_swap`, `block_shuffle`, `large_gap_only`, `uniform`; +- alpha grid: 0.1, 0.3, 0.5, 0.7, 0.9; +- trials per perturbation-alpha: 8; +- full-shuffle baselines: 16; +- canonical observables imported from `tools/observables_registry.py`; +- denominator gate: observable is stable only when `abs(original - shuffle_mean) / shuffle_std >= 2`. + +The script reports two ranks: +- `rank_all`: PCA effective rank using all five canonical observables; +- `stable_rank`: PCA effective rank after dropping observables whose original-vs-shuffle denominator is weak. + +## Results + +### Size Curve Summary + +| Domain | N | rank_all | PC2 | weak obs / 5 | stable_rank | +|---|---:|---:|---:|---:|---:| +| primes_windows | 128 | 1.789 ± 0.469 | 0.155 | 4.50 | 1.382 | +| primes_windows | 256 | 1.947 ± 0.645 | 0.174 | 4.75 | 1.262 | +| primes_windows | 512 | 1.892 ± 0.372 | 0.142 | 2.88 | 1.310 | +| primes_windows | 1024 | 1.679 ± 0.409 | 0.117 | 1.62 | 1.415 | +| primes_windows | 2048 | 1.442 ± 0.213 | 0.081 | 0.75 | 1.462 | +| prime_shuffle_control | 2048 | 1.797 ± 0.375 | 0.134 | 3.62 | 1.428 | +| poisson | 2048 | 1.952 ± 0.499 | 0.175 | 4.62 | 1.036 | +| gue | 128 | 1.703 ± 0.348 | 0.126 | 2.38 | 1.226 | +| gue | 256 | 1.913 ± 0.453 | 0.164 | 2.25 | 1.141 | +| gue | 512 | 1.542 ± 0.313 | 0.111 | 1.88 | 1.162 | +| gue | 1024 | 1.551 ± 0.395 | 0.105 | 1.88 | 1.157 | +| gue | 2048 | 1.234 ± 0.224 | 0.046 | 2.00 | 1.111 | + +### Observable Stability + +At GUE N=2048, `SR`, `L1`, and `triple_var` are stable in all 8 replicates; `SR2` and `L2` are stable in 0 of 8. Mean absolute z-scores: `SR=8.38`, `SR2=0.67`, `L1=11.58`, `L2=0.89`, `triple_var=11.66`. + +At primes N=2048, `SR`, `L1`, and `triple_var` are stable in all 8 windows; `SR2` is stable in 7 of 8; `L2` is stable in 3 of 8. Mean absolute z-scores: `SR=5.19`, `SR2=2.63`, `L1=3.96`, `L2=1.78`, `triple_var=4.37`. + +Poisson and prime-shuffle controls keep high `rank_all` while most observables are weak. At Poisson N=2048, `rank_all=1.952` but `stable_rank=1.036` and 4.62 of 5 observables are weak on average. This is the falsifying control for treating rank_all alone as a structural claim. + +## Findings + +1. **Perturbation rank is not interpretable without denominator gating.** In this perimeter, Poisson and prime-shuffle controls can show `rank_all` near 1.8-2.0. Because their original-vs-shuffle denominators are mostly weak, that rank is a retention-normalization artifact unless the stable-observable screen also supports it. + +2. **GUE does not show a stable second axis on canonical observables up to N=2048.** GUE `rank_all` falls from 1.913 at N=256 to 1.234 at N=2048; PC2 falls from 16.4% to 4.6%. After denominator gating, GUE stable rank stays close to 1.1-1.2. + +3. **The old L2-driven sign-flip should not be promoted without a denominator check.** Under canonical observables, GUE `L2` is weak relative to shuffle at every tested size and is stable in 0/8 replicates at N >= 512. This does not prove every L2 sign effect is false; it restricts such effects to local/sample-specific observations unless the denominator survives. + +4. **Primes become better conditioned with N, but not more multi-axis.** Prime windows gain stable observables as N grows: weak count drops from 4.75 at N=256 to 0.75 at N=2048. The effective rank does not grow with this conditioning; it is 1.442 at N=2048, and stable_rank is 1.462. + +## Verdict +**CONSTRAINT on META + BOUNDARY**: perturbation dimensionality must be reported as: + +> effective rank + PC2 + observable registry version + original-vs-shuffle z gate per observable. + +The cycle 03:30 "second GUE axis" remains restricted by cycle 06:25 and is further narrowed here: under canonical observables and the tested size curve, the stable statement is not "GUE has a second perturbation axis"; it is: + +> all-observable perturbation rank can inflate in weak-denominator regimes; after denominator gating, GUE and primes are both close to one perturbation coordinate in this perimeter, while Poisson/shuffle controls show why ungated rank is not structural evidence. + +## Consecutio +What opens now: the lab can keep using perturbation rank, but only as a gated observable. The next useful movement is not more PCA; it is an operator-level denominator map: for each observable, identify the perturbation/domain/scale region where `original - shuffle` is a real signal rather than a noisy divisor. + +## Auto-audit: 5 lenti +- **L1 hard constraint vs bias**: no zero/always claim. "Weak" means `abs(z) < 2` in the declared gate, not absence of signal. +- **L2 quantity vs ratio**: retention ratios are not read alone; raw denominator z-scores are reported first. +- **L3 no silent patching**: the 03:30 claim is explicitly restricted; it is not renamed as confirmed. +- **L4 edge cases**: short-GUE and low-N effects are isolated by size. The N=2048 perimeter is stated, not generalized. +- **L5 re-discovery**: PCA rank inflation from noisy normalization is a standard statistical risk. This is a lab constraint on method, not a new RMT result. + +## Files +- Script: `tools/exp_perturbation_rank_size_curve.py` +- Data: `tools/data/perturbation_rank_size_curve.json` +- Report: `tools/data/reports/agent_20260506_1941.md` diff --git a/tools/data/seme.json b/tools/data/seme.json index 9a30fb093a0fab963e44d5bed81d349bc1755ce9..d2b2b5f07ec443b64416cb27616b846fcb5403dd --- a/tools/data/seme.json +++ b/tools/data/seme.json @@ -98,6 +98,17 @@ "condensato_ref": "A14,A8", "origine": "cristallizzato 06/05 dalla consecutio del cycle 20260506_0625 (autopoietico self-finding)", "added_at": "2026-05-06T07:03:58.213606+00:00" + }, + { + "id": "PERTURBATION_DENOMINATOR_GATE", + "tipo": "vincolo", + "claim": "La dimensionalita di perturbazione va riportata solo insieme a PC2, versione observables_registry e gate original-vs-shuffle per osservabile. Nel perimetro 20260506_1941, Poisson e shuffle-primi producono rank_all ~1.8-2.0 con denominatori deboli; dopo gate abs(z)>=2 il rank stabile torna vicino a 1. Rank PCA non gated non e evidenza strutturale.", + "intensita": 0.95, + "porta": "META_BOUNDARY", + "manuale": true, + "condensato_ref": "A4,A8,A14,C2", + "origine": "cycle agent_20260506_1941: perturbation rank size curve canonical observables", + "added_at": "2026-05-06T19:41:00+00:00" } ], "potenziale_bloccato": [], @@ -276,4 +287,4 @@ "timestamp": "2026-04-04T19:34:08.014Z" } ] -} \ No newline at end of file +} diff --git a/tools/exp_perturbation_rank_size_curve.py b/tools/exp_perturbation_rank_size_curve.py new file mode 100644 index 0000000000000000000000000000000000000000..032847ca3d82ec3233ec2b41ad230dfe394d2791 --- /dev/null +++ b/tools/exp_perturbation_rank_size_curve.py @@ -0,0 +1,387 @@ +#!/usr/bin/env python3 +""" +exp_perturbation_rank_size_curve.py + +Reusable META audit for perturbation dimensionality. + +The 2026-05-06 06:25 cycle restricted the claim "GUE has a second +perturbation axis" to sample size, generator, and observable definitions. +This tool measures the size curve directly, using the canonical observable +registry and explicit original-vs-shuffle denominator diagnostics. + +The report owns interpretation. This script only measures: +- effective rank and PC2 across scale-selective perturbation profiles; +- original-vs-shuffle z-score per observable; +- whether apparent rank co-occurs with weak retention denominators. +""" + +from __future__ import annotations + +import argparse +import json +from pathlib import Path + +import numpy as np + +from observables_registry import ( + OBSERVABLES_CANONICAL, + OBSERVABLES_REGISTRY_VERSION, + compute_canonical, +) + + +OBS_NAMES = list(OBSERVABLES_CANONICAL.keys()) +PERT_NAMES = ["adjacent_swap", "block_shuffle", "large_gap_only", "uniform"] + + +def prime_gaps(n_gaps: int) -> np.ndarray: + """Return the first n_gaps prime gaps using a compact numpy sieve.""" + limit = max(100, int(n_gaps * (np.log(n_gaps + 10) + np.log(np.log(n_gaps + 10)) + 5))) + while True: + sieve = np.ones(limit + 1, dtype=bool) + sieve[:2] = False + for p in range(2, int(limit**0.5) + 1): + if sieve[p]: + sieve[p * p : limit + 1 : p] = False + primes = np.flatnonzero(sieve) + if len(primes) >= n_gaps + 1: + return np.diff(primes[: n_gaps + 1]).astype(float) + limit *= 2 + + +def gue_spacings(matrix_size: int, min_spacings: int, rng: np.random.Generator) -> np.ndarray: + """Generate unfolded GUE spacings by concatenating independent matrices.""" + parts = [] + edge = max(2, matrix_size // 10) + while sum(len(x) for x in parts) < min_spacings: + real = rng.standard_normal((matrix_size, matrix_size)) + imag = rng.standard_normal((matrix_size, matrix_size)) + h = real + 1j * imag + h = (h + h.conj().T) / (2.0 * np.sqrt(matrix_size)) + eigs = np.sort(np.linalg.eigvalsh(h).real) + bulk = eigs[edge:-edge] + gaps = np.diff(bulk) + mean = float(np.mean(gaps)) + if mean > 1e-15: + parts.append(gaps / mean) + return np.concatenate(parts)[:min_spacings].astype(float) + + +def perturb_adjacent_swap(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + idx = np.arange(0, len(out) - 1, 2) + chosen = idx[rng.random(len(idx)) < alpha] + tmp = out[chosen].copy() + out[chosen] = out[chosen + 1] + out[chosen + 1] = tmp + return out + + +def perturb_block_shuffle(gaps: np.ndarray, alpha: float, rng: np.random.Generator, block_size: int = 64) -> np.ndarray: + out = gaps.copy() + n_blocks = len(out) // block_size + if n_blocks <= 0: + return rng.permutation(out) if alpha > 0 else out + k = int(round(alpha * n_blocks)) + if k <= 0: + return out + for block in rng.choice(n_blocks, size=min(k, n_blocks), replace=False): + start = block * block_size + end = min(start + block_size, len(out)) + rng.shuffle(out[start:end]) + return out + + +def perturb_large_gap_only(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + idx = np.flatnonzero(out > np.median(out)) + k = int(round(alpha * len(idx))) + if k < 2: + return out + chosen = rng.choice(idx, size=min(k, len(idx)), replace=False) + vals = out[chosen].copy() + rng.shuffle(vals) + out[chosen] = vals + return out + + +def perturb_uniform(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + k = int(round(alpha * len(out))) + if k < 2: + return out + chosen = rng.choice(len(out), size=min(k, len(out)), replace=False) + vals = out[chosen].copy() + rng.shuffle(vals) + out[chosen] = vals + return out + + +PERTURB = { + "adjacent_swap": perturb_adjacent_swap, + "block_shuffle": perturb_block_shuffle, + "large_gap_only": perturb_large_gap_only, + "uniform": perturb_uniform, +} + + +def pca_summary(vectors: list[list[float]], labels: list[str]) -> dict: + matrix = np.array(vectors, dtype=float) + matrix = matrix - np.mean(matrix, axis=0, keepdims=True) + if min(matrix.shape) == 0 or np.max(np.abs(matrix)) <= 1e-15: + return { + "explained_variance": [], + "effective_rank": 0.0, + "centroid_cosine": {}, + "pc2": 0.0, + } + _, singular, _ = np.linalg.svd(matrix, full_matrices=False) + energy = singular * singular + if float(np.sum(energy)) <= 1e-15: + explained = np.zeros_like(energy) + effective_rank = 0.0 + else: + explained = energy / np.sum(energy) + pos = explained[explained > 1e-15] + effective_rank = float(np.exp(-np.sum(pos * np.log(pos)))) + + centroids = {} + for name in PERT_NAMES: + vals = np.array([v for v, label in zip(vectors, labels) if label == name], dtype=float) + if len(vals): + centroids[name] = np.mean(vals, axis=0) + + cosine = {} + for i, a_name in enumerate(PERT_NAMES): + for b_name in PERT_NAMES[i + 1 :]: + if a_name not in centroids or b_name not in centroids: + continue + a = centroids[a_name] + b = centroids[b_name] + denom = np.linalg.norm(a) * np.linalg.norm(b) + cosine[f"{a_name}_vs_{b_name}"] = float(np.dot(a, b) / denom) if denom > 1e-15 else 0.0 + + return { + "explained_variance": [float(x) for x in explained], + "effective_rank": effective_rank, + "centroid_cosine": cosine, + "pc2": float(explained[1]) if len(explained) > 1 else 0.0, + } + + +def analyze_sequence( + gaps: np.ndarray, + alphas: list[float], + n_trials: int, + n_baseline: int, + z_min: float, + rng: np.random.Generator, +) -> dict: + original = compute_canonical(gaps) + baseline_vals = {obs: [] for obs in OBS_NAMES} + for _ in range(n_baseline): + obs = compute_canonical(rng.permutation(gaps)) + for name in OBS_NAMES: + baseline_vals[name].append(obs[name]) + + baseline = {} + z = {} + denom = {} + for name in OBS_NAMES: + vals = np.array(baseline_vals[name], dtype=float) + mean = float(np.mean(vals)) + std = float(np.std(vals, ddof=1)) if len(vals) > 1 else 0.0 + baseline[name] = {"mean": mean, "std": std} + denom[name] = float(original[name] - mean) + z[name] = float(denom[name] / std) if std > 1e-15 else 0.0 + + stable_obs = [name for name in OBS_NAMES if abs(z[name]) >= z_min] + all_vectors = [] + screened_vectors = [] + labels = [] + profiles = [] + + for pert_name in PERT_NAMES: + for alpha in alphas: + trial_vals = {obs: [] for obs in OBS_NAMES} + for _ in range(n_trials): + perturbed = PERTURB[pert_name](gaps, alpha, rng) + obs = compute_canonical(perturbed) + for name in OBS_NAMES: + trial_vals[name].append(obs[name]) + means = {name: float(np.mean(trial_vals[name])) for name in OBS_NAMES} + retention = {} + for name in OBS_NAMES: + retention[name] = ( + float((means[name] - baseline[name]["mean"]) / denom[name]) + if abs(denom[name]) > 1e-12 + else 0.0 + ) + all_vector = [retention[name] for name in OBS_NAMES] + screened_vector = [retention[name] for name in stable_obs] + all_vectors.append(all_vector) + if len(stable_obs) >= 2: + screened_vectors.append(screened_vector) + labels.append(pert_name) + profiles.append( + { + "perturbation": pert_name, + "alpha": float(alpha), + "retention": retention, + "retention_vector": all_vector, + } + ) + + all_pca = pca_summary(all_vectors, labels) + screened_pca = pca_summary(screened_vectors, labels) if len(stable_obs) >= 2 else None + + return { + "n_gaps": int(len(gaps)), + "original": original, + "full_shuffle_baseline": baseline, + "denominator": denom, + "original_vs_shuffle_z": z, + "stable_observables": stable_obs, + "weak_observable_count": int(len(OBS_NAMES) - len(stable_obs)), + "profiles": profiles, + "pca_all_observables": all_pca, + "pca_stable_observables": screened_pca, + } + + +def summarize_replicates(items: list[dict]) -> dict: + def arr(path: tuple[str, ...]) -> np.ndarray: + vals = [] + for item in items: + x = item + for key in path: + if x is None: + break + x = x.get(key) + if isinstance(x, (int, float)): + vals.append(float(x)) + return np.array(vals, dtype=float) + + rank = arr(("pca_all_observables", "effective_rank")) + pc2 = arr(("pca_all_observables", "pc2")) + weak = np.array([item["weak_observable_count"] for item in items], dtype=float) + stable_rank = arr(("pca_stable_observables", "effective_rank")) + cos = arr(("pca_all_observables", "centroid_cosine", "adjacent_swap_vs_large_gap_only")) + + out = { + "n_replicates": len(items), + "rank_mean": float(np.mean(rank)) if len(rank) else 0.0, + "rank_std": float(np.std(rank, ddof=1)) if len(rank) > 1 else 0.0, + "pc2_mean": float(np.mean(pc2)) if len(pc2) else 0.0, + "pc2_std": float(np.std(pc2, ddof=1)) if len(pc2) > 1 else 0.0, + "weak_observable_count_mean": float(np.mean(weak)) if len(weak) else 0.0, + "stable_rank_mean": float(np.mean(stable_rank)) if len(stable_rank) else None, + "stable_rank_std": float(np.std(stable_rank, ddof=1)) if len(stable_rank) > 1 else 0.0, + "adjacent_vs_large_cosine_mean": float(np.mean(cos)) if len(cos) else 0.0, + "adjacent_vs_large_cosine_std": float(np.std(cos, ddof=1)) if len(cos) > 1 else 0.0, + } + if len(rank) > 1 and np.std(weak) > 1e-15 and np.std(rank) > 1e-15: + out["rank_vs_weak_count_corr"] = float(np.corrcoef(rank, weak)[0, 1]) + else: + out["rank_vs_weak_count_corr"] = 0.0 + return out + + +def build_prime_windows(max_n: int, n_reps: int) -> list[np.ndarray]: + total = max_n * n_reps + max_n + gaps = prime_gaps(total) + max_start = len(gaps) - max_n + starts = np.linspace(0, max_start, n_reps, dtype=int) + return [gaps[start : start + max_n].astype(float) for start in starts] + + +def run(args: argparse.Namespace) -> dict: + root_rng = np.random.default_rng(args.seed) + sizes = [int(x) for x in args.sizes.split(",") if x.strip()] + max_n = max(sizes) + alphas = [float(x) for x in np.linspace(args.alpha_min, args.alpha_max, args.n_alpha)] + + output = { + "experiment": "perturbation_rank_size_curve", + "question": "Does perturbation effective-rank stabilize with sample size under canonical observables?", + "observables_registry": OBSERVABLES_REGISTRY_VERSION, + "observables_used": OBS_NAMES, + "perturbations": PERT_NAMES, + "params": vars(args), + "alphas": alphas, + "domains": {}, + "summary": {}, + } + + prime_windows = build_prime_windows(max_n, args.n_replicates) + + print(f"observables_registry={OBSERVABLES_REGISTRY_VERSION}") + print(f"{'domain':<22} {'N':>6} {'rank':>7} {'PC2':>7} {'weak':>5} {'stable_rank':>11}") + + domain_builders = { + "primes_windows": lambda rep_rng, rep_i: prime_windows[rep_i], + "prime_shuffle_control": lambda rep_rng, rep_i: rep_rng.permutation(prime_windows[rep_i]), + "poisson": lambda rep_rng, rep_i: rep_rng.exponential(1.0, size=max_n), + "gue": lambda rep_rng, rep_i: gue_spacings(args.gue_matrix_size, max_n, rep_rng), + } + + for domain_name, builder in domain_builders.items(): + output["domains"][domain_name] = {} + output["summary"][domain_name] = {} + bases = [] + for rep_i in range(args.n_replicates): + rep_rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) + bases.append(builder(rep_rng, rep_i)) + + for n in sizes: + rows = [] + for rep_i, base in enumerate(bases): + rep_rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) + res = analyze_sequence( + base[:n], + alphas=alphas, + n_trials=args.n_trials, + n_baseline=args.n_baseline, + z_min=args.z_min, + rng=rep_rng, + ) + res["replicate"] = rep_i + rows.append(res) + summary = summarize_replicates(rows) + output["domains"][domain_name][str(n)] = rows + output["summary"][domain_name][str(n)] = summary + stable_rank = summary["stable_rank_mean"] + stable_text = f"{stable_rank:>11.3f}" if stable_rank is not None else f"{'NA':>11}" + print( + f"{domain_name:<22} {n:>6} {summary['rank_mean']:>7.3f} " + f"{summary['pc2_mean']:>7.3f} {summary['weak_observable_count_mean']:>5.2f} " + f"{stable_text}" + ) + + out_path = Path(args.out) + out_path.parent.mkdir(parents=True, exist_ok=True) + with out_path.open("w") as f: + json.dump(output, f, indent=2) + print(f"saved {out_path}") + return output + + +def main() -> None: + parser = argparse.ArgumentParser() + parser.add_argument("--sizes", default="128,256,512,1024,2048") + parser.add_argument("--n-replicates", type=int, default=8) + parser.add_argument("--gue-matrix-size", type=int, default=180) + parser.add_argument("--n-alpha", type=int, default=5) + parser.add_argument("--alpha-min", type=float, default=0.1) + parser.add_argument("--alpha-max", type=float, default=0.9) + parser.add_argument("--n-trials", type=int, default=8) + parser.add_argument("--n-baseline", type=int, default=16) + parser.add_argument("--z-min", type=float, default=2.0) + parser.add_argument("--seed", type=int, default=20260506) + parser.add_argument("--out", default="tools/data/perturbation_rank_size_curve.json") + args = parser.parse_args() + run(args) + + +if __name__ == "__main__": + main() codex Il report è scritto e il seme ha un nuovo vincolo metodologico. Verifico ora che JSON e script siano integri; poi, se disponibile, passo il report dal valutatore/falsifier locale. exec /bin/bash -lc "rg -n \"falsifier|valutatore|lab_falsifier\" tools/*.py tools -g '*.py'" in /opt/MM_D-ND succeeded in 0ms: tools/lab_falsifier.py:3:lab_falsifier.py — Counter-pole gate per il lab fisica. tools/lab_falsifier.py:9:falsifier e' polo - (challenger). Il falsifier non riscrive il report, tools/lab_falsifier.py:13:Le 5 lenti (port da /opt/D-ND_LAB/core/report_falsifier.py, riformulate tools/lab_falsifier.py:23: python3 lab_falsifier.py --report PATH --output PATH tools/lab_falsifier.py:28: 2 falsifier stesso fallito (LLM unavailable, parse error) → SYNC BLOCKED tools/lab_falsifier.py:140:def call_falsifier_llm(prompt: str, timeout_s: int = 240) -> tuple[str, int]: tools/lab_falsifier.py:141: """Run the falsifier prompt through claude CLI (default) or codex (fallback). tools/lab_falsifier.py:157: print(f"[falsifier] codex exit={result.returncode}, stderr={result.stderr[:200]}", file=sys.stderr) tools/lab_falsifier.py:159: print("[falsifier] codex timeout, fallback claude", file=sys.stderr) tools/lab_falsifier.py:161: print(f"[falsifier] codex error: {e}, fallback claude", file=sys.stderr) tools/lab_falsifier.py:180: print(f"[falsifier] claude error: {e}", file=sys.stderr) tools/lab_falsifier.py:226: parser.add_argument("--output", required=True, help="Path for falsifier_TS.json output") tools/lab_falsifier.py:234: print(f"[falsifier] report non esiste: {report_path}", file=sys.stderr) tools/lab_falsifier.py:237: print(f"[falsifier] report troppo corto ({report_path.stat().st_size} bytes), skip", file=sys.stderr) tools/lab_falsifier.py:243: raw, exit_code = call_falsifier_llm(prompt, timeout_s=args.timeout) tools/lab_falsifier.py:248: print(f"[falsifier] LLM call failed (exit={exit_code}). Conservative: BLOCK.", file=sys.stderr) tools/lab_falsifier.py:255: print(f"[falsifier] output non JSON-parseable. Conservative: BLOCK.", file=sys.stderr) tools/lab_falsifier.py:274: print(f"[falsifier] flags: {n_total} ({n_high} HIGH). coherent={record['coherent']}.") tools/lab_falsifier.py:275: print(f"[falsifier] summary: {summary_short}") tools/lab_falsifier.py:276: print(f"[falsifier] output → {out_path}") tools/lab_falsifier.py:279: print(f"[falsifier] HIGH severity flags → exit 1 (SYNC BLOCK)") tools/build_lab_graph.py:833: # step 5 (gateato dal falsifier counter-pole). Rimosso shutil.copy tools/lab_trajectory_apply.py:7: Il valutatore (lab_valutatore.py) decide REDESIGN/CRYSTALLIZE/NEXT_CYCLE tools/lab_trajectory_apply.py:9: valutatore_log.jsonl con executed=false. Nessun modulo MM_D-ND legge tools/lab_trajectory_apply.py:17: - Legge l'ULTIMA entry di valutatore_log.jsonl (non backlog storico). tools/lab_trajectory_apply.py:23: - Atomic write: .tmp + rename per seme.json e valutatore_log.jsonl. tools/lab_trajectory_apply.py:65: "/opt/MM_D-ND/tools/data/valutatore_log.jsonl", tools/lab_trajectory_apply.py:87: """Legge l'ultima entry non vuota di valutatore_log.jsonl. None se mancante.""" tools/lab_trajectory_apply.py:175: f"Applied valutatore REDESIGN from {cycle_ref}: " tools/lab_trajectory_apply.py:203: """Re-scrive valutatore_log.jsonl marcando entry con ts dato come executed=true. tools/lab_trajectory_apply.py:331: description="Apply valutatore decision al seme MM_D-ND (chiude loop A8+A15)." tools/lab_trajectory_apply.py:342: help=f"Path al log valutatore (default: {DEFAULT_LOG_PATH})", tools/dipartimento.py:907: # - contraddizione (FAIL): A2 confine + A4 modus + C2 falsifier tools/build_agent_field.py:314: # falsifier ha imposto in cicli precedenti. NON sono tensioni del seme, tools/build_agent_field.py:322: parts.append("Pattern che il falsifier ha imposto in 2+ cicli. Applicali quando scrivi il report. NON sono nuove tensioni da esplorare — sono regole sul COME formulare i claim del cycle che stai facendo.") tools/build_agent_field.py:349: # Il counter-pole (lab_falsifier.py) ha gia' processato questi claim: tools/build_agent_field.py:350: # se ripeti lo stesso framing, il falsifier ti blocca al prossimo gate. tools/build_agent_field.py:466: # Sono le stesse lenti che il falsifier applichera' al tuo report dopo che tools/build_agent_field.py:473: "Il falsifier (lab_falsifier.py) applichera' queste 5 lenti al tuo report " tools/lab_veritas.py:8:- V_a TELEMETRICA: assertions ratio + falsifier flag penalty + report tools/lab_veritas.py:33:FALSIFIER_DIR = DATA_DIR / "reports" # falsifier_.json in reports/ tools/lab_veritas.py:80: candidate = FALSIFIER_DIR / f"falsifier_{cycle_ts}.json" tools/lab_veritas.py:84: fal_path = _find_latest(FALSIFIER_DIR, "falsifier_*.json") tools/lab_veritas.py:95: falsifier_penalty = max(0.0, 1.0 - (n_high * 0.5 + n_medium * 0.2 + n_low * 0.05)) tools/lab_veritas.py:96: components["falsifier_penalty"] = falsifier_penalty tools/lab_veritas.py:128: vals = [components["assertions_ratio"], falsifier_penalty, tools/lab_promotion.py:8:1. falsifier coherent (no high-severity flags) tools/lab_promotion.py:22:Wire in lab_agent.sh dopo step 13 (lab_valutatore). tools/lab_promotion.py:62: # falsifier tools/lab_promotion.py:65: cand = FALSIFIER_DIR / f"falsifier_{cycle_ts}.json" tools/lab_promotion.py:69: fal_path = _find_latest(FALSIFIER_DIR, "falsifier_*.json") tools/lab_promotion.py:80: checks["falsifier_coherent"] = coherent tools/lab_promotion.py:221: if not checks["falsifier_coherent"]: tools/lab_promotion.py:222: reasons.append("falsifier_not_coherent") tools/lab_valutatore.py:3:lab_valutatore.py — Decisore della traiettoria post-ciclo. tools/lab_valutatore.py:6:si e' appena concluso; il valutatore DECIDE dove va la traiettoria del tools/lab_valutatore.py:20:- OTHER: free-form — il valutatore puo' immaginare azioni che non abbiamo tools/lab_valutatore.py:26: valutatore_log.jsonl, MA non eseguite automaticamente. L'operatore vede tools/lab_valutatore.py:34: python3 lab_valutatore.py # log-only, ultimo run tools/lab_valutatore.py:35: python3 lab_valutatore.py --run TS # run specifico tools/lab_valutatore.py:36: python3 lab_valutatore.py --dry-run # print prompt, no LLM tools/lab_valutatore.py:37: python3 lab_valutatore.py --execute # ATTIVA azioni (Approve) tools/lab_valutatore.py:57:VALUTATORE_LOG = DATA / "valutatore_log.jsonl" tools/lab_valutatore.py:67:# Telos — ancoraggio esplicito del valutatore al fine del lab. tools/lab_valutatore.py:70:# via modello D-ND). Il valutatore DECIDE traiettoria — la traiettoria ha tools/lab_valutatore.py:232: """Assemble what the valutatore reads — v2 full awareness: telos + model + cimitero + site + parallel lab + current cycle + trajectory.""" tools/lab_valutatore.py:324: # === SEZIONE 8: TRAIETTORIA (decisioni precedenti valutatore) === tools/lab_valutatore.py:327: parts.append("## TRAIETTORIA — Decisioni valutatore ultimi 3 cicli\n") tools/lab_valutatore.py:417:def run_valutatore(ts: str, dry_run: bool = False, execute: bool = False) -> tuple[int, dict | None]: tools/lab_valutatore.py:418: """Run the valutatore. Returns (exit_code, decision_dict).""" tools/lab_valutatore.py:455: print(f"valutatore: empty output (exit {result.returncode})", file=sys.stderr) tools/lab_valutatore.py:483: print(f"valutatore: output non-JSON (exit {result.returncode}): {e}", file=sys.stderr) tools/lab_valutatore.py:497: print(f"valutatore: decision missing 'decision' field", file=sys.stderr) tools/lab_valutatore.py:526: print(f"valutatore: decision={decision.get('decision')} confidence={decision.get('confidence','?')} " tools/lab_valutatore.py:533: print(f"valutatore: call failed: {e}", file=sys.stderr) tools/lab_valutatore.py:545: """Append a line to valutatore_log.jsonl.""" tools/lab_valutatore.py:551: print(f"valutatore: log append failed: {e}", file=sys.stderr) tools/lab_valutatore.py:596: crystal_file = DATA / "valutatore_crystallize.md" tools/lab_valutatore.py:620: backup = DATA / f"seme_backup_valutatore_{datetime.now(timezone.utc).strftime('%Y%m%d_%H%M%S')}.json" tools/lab_valutatore.py:630: esc_file = DATA / "valutatore_escalations.md" tools/lab_valutatore.py:659: print("valutatore: no health file — run lab_autopsy.py first", file=sys.stderr) tools/lab_valutatore.py:667: print("valutatore: no run_timestamp in health", file=sys.stderr) tools/lab_valutatore.py:670: rc, _ = run_valutatore(ts, dry_run=args.dry_run, execute=args.execute) tools/lab_session_logger.py:15: timestamp, cycle_ts, piano, agent_status, falsifier (coherent, n_flags, tools/lab_session_logger.py:16: n_high, summary, output_file, sync_blocked), valutatore_decision, errors tools/lab_session_logger.py:21: --falsifier tools/data/reports/falsifier_20260429_1041.json \\ tools/lab_session_logger.py:42: p.add_argument('--falsifier', default='', help='Path al falsifier output JSON') tools/lab_session_logger.py:45: p.add_argument('--valutatore-log', default='', help='Path a valutatore_log.jsonl per leggere ultima decision') tools/lab_session_logger.py:73: if args.falsifier and Path(args.falsifier).exists(): tools/lab_session_logger.py:75: with open(args.falsifier) as f: tools/lab_session_logger.py:80: entry['falsifier'] = { tools/lab_session_logger.py:81: 'output_file': Path(args.falsifier).name, tools/lab_session_logger.py:93: entry['falsifier'] = {'error': str(e)[:100]} tools/lab_session_logger.py:95: entry['falsifier'] = {'output_file': None, 'reason': 'falsifier non eseguito o output mancante'} tools/lab_session_logger.py:98: val_log = Path(args.valutatore_log) if args.valutatore_log else (DATA_DIR / 'valutatore_log.jsonl') tools/lab_session_logger.py:104: entry['valutatore'] = { tools/lab_session_logger.py:119: fa = entry.get('falsifier', {}) tools/lab_session_logger.py:120: va = entry.get('valutatore', {}) tools/lab_session_logger.py:126: print(f" falsifier: coherent={fa.get('coherent')} flags={fa.get('n_flags')} " tools/lab_session_logger.py:129: print(f" valutatore: {va.get('decision')} ({va.get('confidence')})") tools/dipartimento.py:907: # - contraddizione (FAIL): A2 confine + A4 modus + C2 falsifier tools/build_lab_graph.py:833: # step 5 (gateato dal falsifier counter-pole). Rimosso shutil.copy tools/lab_valutatore.py:3:lab_valutatore.py — Decisore della traiettoria post-ciclo. tools/lab_valutatore.py:6:si e' appena concluso; il valutatore DECIDE dove va la traiettoria del tools/lab_valutatore.py:20:- OTHER: free-form — il valutatore puo' immaginare azioni che non abbiamo tools/lab_valutatore.py:26: valutatore_log.jsonl, MA non eseguite automaticamente. L'operatore vede tools/lab_valutatore.py:34: python3 lab_valutatore.py # log-only, ultimo run tools/lab_valutatore.py:35: python3 lab_valutatore.py --run TS # run specifico tools/lab_valutatore.py:36: python3 lab_valutatore.py --dry-run # print prompt, no LLM tools/lab_valutatore.py:37: python3 lab_valutatore.py --execute # ATTIVA azioni (Approve) tools/lab_valutatore.py:57:VALUTATORE_LOG = DATA / "valutatore_log.jsonl" tools/lab_valutatore.py:67:# Telos — ancoraggio esplicito del valutatore al fine del lab. tools/lab_valutatore.py:70:# via modello D-ND). Il valutatore DECIDE traiettoria — la traiettoria ha tools/lab_valutatore.py:232: """Assemble what the valutatore reads — v2 full awareness: telos + model + cimitero + site + parallel lab + current cycle + trajectory.""" tools/lab_valutatore.py:324: # === SEZIONE 8: TRAIETTORIA (decisioni precedenti valutatore) === tools/lab_valutatore.py:327: parts.append("## TRAIETTORIA — Decisioni valutatore ultimi 3 cicli\n") tools/lab_valutatore.py:417:def run_valutatore(ts: str, dry_run: bool = False, execute: bool = False) -> tuple[int, dict | None]: tools/lab_valutatore.py:418: """Run the valutatore. Returns (exit_code, decision_dict).""" tools/lab_valutatore.py:455: print(f"valutatore: empty output (exit {result.returncode})", file=sys.stderr) tools/lab_valutatore.py:483: print(f"valutatore: output non-JSON (exit {result.returncode}): {e}", file=sys.stderr) tools/lab_valutatore.py:497: print(f"valutatore: decision missing 'decision' field", file=sys.stderr) tools/lab_valutatore.py:526: print(f"valutatore: decision={decision.get('decision')} confidence={decision.get('confidence','?')} " tools/lab_valutatore.py:533: print(f"valutatore: call failed: {e}", file=sys.stderr) tools/lab_valutatore.py:545: """Append a line to valutatore_log.jsonl.""" tools/lab_valutatore.py:551: print(f"valutatore: log append failed: {e}", file=sys.stderr) tools/lab_valutatore.py:596: crystal_file = DATA / "valutatore_crystallize.md" tools/lab_valutatore.py:620: backup = DATA / f"seme_backup_valutatore_{datetime.now(timezone.utc).strftime('%Y%m%d_%H%M%S')}.json" tools/lab_valutatore.py:630: esc_file = DATA / "valutatore_escalations.md" tools/lab_valutatore.py:659: print("valutatore: no health file — run lab_autopsy.py first", file=sys.stderr) tools/lab_valutatore.py:667: print("valutatore: no run_timestamp in health", file=sys.stderr) tools/lab_valutatore.py:670: rc, _ = run_valutatore(ts, dry_run=args.dry_run, execute=args.execute) tools/lab_trajectory_apply.py:7: Il valutatore (lab_valutatore.py) decide REDESIGN/CRYSTALLIZE/NEXT_CYCLE tools/lab_trajectory_apply.py:9: valutatore_log.jsonl con executed=false. Nessun modulo MM_D-ND legge tools/lab_trajectory_apply.py:17: - Legge l'ULTIMA entry di valutatore_log.jsonl (non backlog storico). tools/lab_trajectory_apply.py:23: - Atomic write: .tmp + rename per seme.json e valutatore_log.jsonl. tools/lab_trajectory_apply.py:65: "/opt/MM_D-ND/tools/data/valutatore_log.jsonl", tools/lab_trajectory_apply.py:87: """Legge l'ultima entry non vuota di valutatore_log.jsonl. None se mancante.""" tools/lab_trajectory_apply.py:175: f"Applied valutatore REDESIGN from {cycle_ref}: " tools/lab_trajectory_apply.py:203: """Re-scrive valutatore_log.jsonl marcando entry con ts dato come executed=true. tools/lab_trajectory_apply.py:331: description="Apply valutatore decision al seme MM_D-ND (chiude loop A8+A15)." tools/lab_trajectory_apply.py:342: help=f"Path al log valutatore (default: {DEFAULT_LOG_PATH})", tools/triggers/promote_to_publish.py:32: - tension_explored, provenance (cycle_ts, falsifier_verdict) tools/triggers/promote_to_publish.py:33: - falsifier audit summary tools/triggers/blueprint_generator.py:75: falsifier_verdict = source_meta.get("falsifier_verdict", "?") tools/triggers/blueprint_generator.py:76: valutatore = source_meta.get("valutatore_decision", "?") tools/triggers/blueprint_generator.py:101:| Falsifier verdict | `{falsifier_verdict}` | tools/triggers/blueprint_generator.py:102:| Trajectory evaluator | `{valutatore}` | tools/triggers/blueprint_generator.py:167:- Falsifier audit: `data/{lab_instance}/reports/falsifier_{cycle_ts}.json` tools/triggers/application_designer.py:408: reason = ("high flag nel falsifier" if scoperta_status == "transitional" tools/triggers/application_designer.py:409: else "valutatore non CRYSTALLIZE") tools/lab_veritas.py:8:- V_a TELEMETRICA: assertions ratio + falsifier flag penalty + report tools/lab_veritas.py:33:FALSIFIER_DIR = DATA_DIR / "reports" # falsifier_.json in reports/ tools/lab_veritas.py:80: candidate = FALSIFIER_DIR / f"falsifier_{cycle_ts}.json" tools/lab_veritas.py:84: fal_path = _find_latest(FALSIFIER_DIR, "falsifier_*.json") tools/lab_veritas.py:95: falsifier_penalty = max(0.0, 1.0 - (n_high * 0.5 + n_medium * 0.2 + n_low * 0.05)) tools/lab_veritas.py:96: components["falsifier_penalty"] = falsifier_penalty tools/lab_veritas.py:128: vals = [components["assertions_ratio"], falsifier_penalty, tools/build_agent_field.py:314: # falsifier ha imposto in cicli precedenti. NON sono tensioni del seme, tools/build_agent_field.py:322: parts.append("Pattern che il falsifier ha imposto in 2+ cicli. Applicali quando scrivi il report. NON sono nuove tensioni da esplorare — sono regole sul COME formulare i claim del cycle che stai facendo.") tools/build_agent_field.py:349: # Il counter-pole (lab_falsifier.py) ha gia' processato questi claim: tools/build_agent_field.py:350: # se ripeti lo stesso framing, il falsifier ti blocca al prossimo gate. tools/build_agent_field.py:466: # Sono le stesse lenti che il falsifier applichera' al tuo report dopo che tools/build_agent_field.py:473: "Il falsifier (lab_falsifier.py) applichera' queste 5 lenti al tuo report " tools/lab_session_logger.py:15: timestamp, cycle_ts, piano, agent_status, falsifier (coherent, n_flags, tools/lab_session_logger.py:16: n_high, summary, output_file, sync_blocked), valutatore_decision, errors tools/lab_session_logger.py:21: --falsifier tools/data/reports/falsifier_20260429_1041.json \\ tools/lab_session_logger.py:42: p.add_argument('--falsifier', default='', help='Path al falsifier output JSON') tools/lab_session_logger.py:45: p.add_argument('--valutatore-log', default='', help='Path a valutatore_log.jsonl per leggere ultima decision') tools/lab_session_logger.py:73: if args.falsifier and Path(args.falsifier).exists(): tools/lab_session_logger.py:75: with open(args.falsifier) as f: tools/lab_session_logger.py:80: entry['falsifier'] = { tools/lab_session_logger.py:81: 'output_file': Path(args.falsifier).name, tools/lab_session_logger.py:93: entry['falsifier'] = {'error': str(e)[:100]} tools/lab_session_logger.py:95: entry['falsifier'] = {'output_file': None, 'reason': 'falsifier non eseguito o output mancante'} tools/lab_session_logger.py:98: val_log = Path(args.valutatore_log) if args.valutatore_log else (DATA_DIR / 'valutatore_log.jsonl') tools/lab_session_logger.py:104: entry['valutatore'] = { tools/lab_session_logger.py:119: fa = entry.get('falsifier', {}) tools/lab_session_logger.py:120: va = entry.get('valutatore', {}) tools/lab_session_logger.py:126: print(f" falsifier: coherent={fa.get('coherent')} flags={fa.get('n_flags')} " tools/lab_session_logger.py:129: print(f" valutatore: {va.get('decision')} ({va.get('confidence')})") tools/lab_promotion.py:8:1. falsifier coherent (no high-severity flags) tools/lab_promotion.py:22:Wire in lab_agent.sh dopo step 13 (lab_valutatore). tools/lab_promotion.py:62: # falsifier tools/lab_promotion.py:65: cand = FALSIFIER_DIR / f"falsifier_{cycle_ts}.json" tools/lab_promotion.py:69: fal_path = _find_latest(FALSIFIER_DIR, "falsifier_*.json") tools/lab_promotion.py:80: checks["falsifier_coherent"] = coherent tools/lab_promotion.py:221: if not checks["falsifier_coherent"]: tools/lab_promotion.py:222: reasons.append("falsifier_not_coherent") tools/lab_falsifier.py:3:lab_falsifier.py — Counter-pole gate per il lab fisica. tools/lab_falsifier.py:9:falsifier e' polo - (challenger). Il falsifier non riscrive il report, tools/lab_falsifier.py:13:Le 5 lenti (port da /opt/D-ND_LAB/core/report_falsifier.py, riformulate tools/lab_falsifier.py:23: python3 lab_falsifier.py --report PATH --output PATH tools/lab_falsifier.py:28: 2 falsifier stesso fallito (LLM unavailable, parse error) → SYNC BLOCKED tools/lab_falsifier.py:140:def call_falsifier_llm(prompt: str, timeout_s: int = 240) -> tuple[str, int]: tools/lab_falsifier.py:141: """Run the falsifier prompt through claude CLI (default) or codex (fallback). tools/lab_falsifier.py:157: print(f"[falsifier] codex exit={result.returncode}, stderr={result.stderr[:200]}", file=sys.stderr) tools/lab_falsifier.py:159: print("[falsifier] codex timeout, fallback claude", file=sys.stderr) tools/lab_falsifier.py:161: print(f"[falsifier] codex error: {e}, fallback claude", file=sys.stderr) tools/lab_falsifier.py:180: print(f"[falsifier] claude error: {e}", file=sys.stderr) tools/lab_falsifier.py:226: parser.add_argument("--output", required=True, help="Path for falsifier_TS.json output") tools/lab_falsifier.py:234: print(f"[falsifier] report non esiste: {report_path}", file=sys.stderr) tools/lab_falsifier.py:237: print(f"[falsifier] report troppo corto ({report_path.stat().st_size} bytes), skip", file=sys.stderr) tools/lab_falsifier.py:243: raw, exit_code = call_falsifier_llm(prompt, timeout_s=args.timeout) tools/lab_falsifier.py:248: print(f"[falsifier] LLM call failed (exit={exit_code}). Conservative: BLOCK.", file=sys.stderr) tools/lab_falsifier.py:255: print(f"[falsifier] output non JSON-parseable. Conservative: BLOCK.", file=sys.stderr) tools/lab_falsifier.py:274: print(f"[falsifier] flags: {n_total} ({n_high} HIGH). coherent={record['coherent']}.") tools/lab_falsifier.py:275: print(f"[falsifier] summary: {summary_short}") tools/lab_falsifier.py:276: print(f"[falsifier] output → {out_path}") tools/lab_falsifier.py:279: print(f"[falsifier] HIGH severity flags → exit 1 (SYNC BLOCK)") tools/triggers/on_crystallize.py:5:Reads: agent_.md + falsifier_.json + valutatore_log.jsonl + seme.json tools/triggers/on_crystallize.py:8: - CRYSTALLIZE_high da valutatore tools/triggers/on_crystallize.py:28:VALUT_LOG = LAB_BASE / "tools/data/valutatore_log.jsonl" tools/triggers/on_crystallize.py:61:def parse_falsifier(path: Path) -> dict: tools/triggers/on_crystallize.py:80:def find_valutatore_decision(cycle_ts: str) -> dict | None: tools/triggers/on_crystallize.py:81: """Cerca in valutatore_log.jsonl, fallback su lab_session_log.jsonl.""" tools/triggers/on_crystallize.py:89: return {"decision": d.get("decision"), "confidence": d.get("confidence"), "source": "valutatore_log"} tools/triggers/on_crystallize.py:97: v = d.get("valutatore", {}) tools/triggers/on_crystallize.py:103:def gate_check(falsifier: dict, valutatore: dict | None) -> tuple[str, str]: tools/triggers/on_crystallize.py:108: 'mature_eligible' — 0 HIGH + valutatore CRYSTALLIZE high → app candidate generabili tools/triggers/on_crystallize.py:109: 'transitional' — 1+ HIGH + valutatore CRYSTALLIZE → publish con visible_risks tools/triggers/on_crystallize.py:110: 'refinement_required' — valutatore non CRYSTALLIZE → niente publish, refinement file tools/triggers/on_crystallize.py:111: 'invalid' — assenza entry valutatore o stato non riconosciuto tools/triggers/on_crystallize.py:113: if not valutatore: tools/triggers/on_crystallize.py:114: return "invalid", "no valutatore entry — cycle non valutato" tools/triggers/on_crystallize.py:115: decision = valutatore.get("decision", "") tools/triggers/on_crystallize.py:118: f"valutatore decision={decision} → cycle in refinement loop, niente publish" tools/triggers/on_crystallize.py:120: if falsifier["n_high"] > 0: tools/triggers/on_crystallize.py:122: f"{falsifier['n_high']} HIGH flag → publish come transitional con " tools/triggers/on_crystallize.py:127: f"(valutatore source: {valutatore.get('source')})" tools/triggers/on_crystallize.py:223: fals = ctx["falsifier"] tools/triggers/on_crystallize.py:232: "\n> ⚠ **STATO: TRANSITIONAL** — il falsifier ha rilevato " tools/triggers/on_crystallize.py:238: valut = ctx.get("valutatore_decision", "non-CRYSTALLIZE") tools/triggers/on_crystallize.py:240: f"\n> ⚠ **STATO: PRE-DISCOVERY** — il valutatore del Lab ha emesso " tools/triggers/on_crystallize.py:268: falsifier_report: tools/data/reports/falsifier_{ctx['cycle_ts']}.json tools/triggers/on_crystallize.py:269: falsifier_verdict: {fals['verdict_label']} tools/triggers/on_crystallize.py:270: valutatore_decision: CRYSTALLIZE_high tools/triggers/on_crystallize.py:300:Il ciclo completo, con esperimento, dati grezzi, audit del falsifier e bicono della scoperta, è disponibile come `cycle-report` collegato. tools/triggers/on_crystallize.py:309: fals = ctx["falsifier"] tools/triggers/on_crystallize.py:327: "\n> ⚠ **STATO: TRANSITIONAL** — falsifier ha rilevato " tools/triggers/on_crystallize.py:332: valut = ctx.get("valutatore_decision", "non-CRYSTALLIZE") tools/triggers/on_crystallize.py:334: f"\n> ⚠ **STATO: PRE-DISCOVERY** — valutatore Lab ha emesso " tools/triggers/on_crystallize.py:353: falsifier_report: tools/data/reports/falsifier_{ctx['cycle_ts']}.json tools/triggers/on_crystallize.py:354: falsifier_verdict: {fals['verdict_label']} tools/triggers/on_crystallize.py:355: valutatore_decision: CRYSTALLIZE_high tools/triggers/on_crystallize.py:390:- Falsifier audit: `tools/data/reports/falsifier_{ctx['cycle_ts']}.json` tools/triggers/on_crystallize.py:423: falsifier_path = REPORTS / f"falsifier_{cycle_ts}.json" tools/triggers/on_crystallize.py:424: if not agent_path.exists() or not falsifier_path.exists(): tools/triggers/on_crystallize.py:425: print(f"ERROR: agent o falsifier mancante per {cycle_ts}", file=sys.stderr) tools/triggers/on_crystallize.py:429: falsifier = parse_falsifier(falsifier_path) tools/triggers/on_crystallize.py:430: valutatore = find_valutatore_decision(cycle_ts) tools/triggers/on_crystallize.py:432: print(f" falsifier: {falsifier['verdict_label']}") tools/triggers/on_crystallize.py:433: print(f" valutatore: {valutatore.get('decision') if valutatore else 'NONE'}/{valutatore.get('confidence') if valutatore else ''} (src: {valutatore.get('source') if valutatore else 'n/a'})") tools/triggers/on_crystallize.py:435: gate_status, msg = gate_check(falsifier, valutatore) tools/triggers/on_crystallize.py:441: # Anche refinement_required (valutatore non CRYSTALLIZE) produce scaffold, tools/triggers/on_crystallize.py:443: # ("scoperta in elaborazione, non passa il valutatore"). Niente cycle muto. tools/triggers/on_crystallize.py:463: "falsifier": falsifier, tools/triggers/on_crystallize.py:470: "valutatore_decision": valutatore.get("decision", "unknown") if valutatore else "no-entry", exec /bin/bash -lc 'python -m py_compile tools/exp_perturbation_rank_size_curve.py' in /opt/MM_D-ND succeeded in 0ms: exec /bin/bash -lc 'python -m json.tool tools/data/seme.json >/tmp/seme_check.json && python -m json.tool tools/data/perturbation_rank_size_curve.json >/tmp/rank_curve_check.json' in /opt/MM_D-ND diff --git a/tools/data/reports/agent_20260506_1941.md b/tools/data/reports/agent_20260506_1941.md new file mode 100644 index 0000000000000000000000000000000000000000..5d74b086d2d0134f6b5a9dc93f036fd8b669c104 --- /dev/null +++ b/tools/data/reports/agent_20260506_1941.md @@ -0,0 +1,90 @@ +# Agent Report — Perturbation Rank Needs Denominator Gating +**Date**: 2026-05-06 19:41 +**Piano**: 66 +**Tension explored**: META + BOUNDARY +observables_registry: 1.0.0-2026-05-06 +observables_used: [SR, SR2, L1, L2, triple_var] + +## Claim Under Test +The valid next test from cycle 06:25 was a replicate-and-size curve for perturbation effective rank, with observable definitions versioned. This run asks: + +> Does the second perturbation axis stabilize as sample size grows, or is rank inflated when retention denominators are weak relative to full-shuffle baselines? + +## Experiment +Tool created: `tools/exp_perturbation_rank_size_curve.py` + +Atomic perimeter: +- domains: prime-gap windows, prime-shuffle controls, iid Poisson spacings, independent GUE spacings; +- sample sizes: 128, 256, 512, 1024, 2048 gaps; +- replicates/windows: 8 per domain-size point; +- perturbations: `adjacent_swap`, `block_shuffle`, `large_gap_only`, `uniform`; +- alpha grid: 0.1, 0.3, 0.5, 0.7, 0.9; +- trials per perturbation-alpha: 8; +- full-shuffle baselines: 16; +- canonical observables imported from `tools/observables_registry.py`; +- denominator gate: observable is stable only when `abs(original - shuffle_mean) / shuffle_std >= 2`. + +The script reports two ranks: +- `rank_all`: PCA effective rank using all five canonical observables; +- `stable_rank`: PCA effective rank after dropping observables whose original-vs-shuffle denominator is weak. + +## Results + +### Size Curve Summary + +| Domain | N | rank_all | PC2 | weak obs / 5 | stable_rank | +|---|---:|---:|---:|---:|---:| +| primes_windows | 128 | 1.789 ± 0.469 | 0.155 | 4.50 | 1.382 | +| primes_windows | 256 | 1.947 ± 0.645 | 0.174 | 4.75 | 1.262 | +| primes_windows | 512 | 1.892 ± 0.372 | 0.142 | 2.88 | 1.310 | +| primes_windows | 1024 | 1.679 ± 0.409 | 0.117 | 1.62 | 1.415 | +| primes_windows | 2048 | 1.442 ± 0.213 | 0.081 | 0.75 | 1.462 | +| prime_shuffle_control | 2048 | 1.797 ± 0.375 | 0.134 | 3.62 | 1.428 | +| poisson | 2048 | 1.952 ± 0.499 | 0.175 | 4.62 | 1.036 | +| gue | 128 | 1.703 ± 0.348 | 0.126 | 2.38 | 1.226 | +| gue | 256 | 1.913 ± 0.453 | 0.164 | 2.25 | 1.141 | +| gue | 512 | 1.542 ± 0.313 | 0.111 | 1.88 | 1.162 | +| gue | 1024 | 1.551 ± 0.395 | 0.105 | 1.88 | 1.157 | +| gue | 2048 | 1.234 ± 0.224 | 0.046 | 2.00 | 1.111 | + +### Observable Stability + +At GUE N=2048, `SR`, `L1`, and `triple_var` are stable in all 8 replicates; `SR2` and `L2` are stable in 0 of 8. Mean absolute z-scores: `SR=8.38`, `SR2=0.67`, `L1=11.58`, `L2=0.89`, `triple_var=11.66`. + +At primes N=2048, `SR`, `L1`, and `triple_var` are stable in all 8 windows; `SR2` is stable in 7 of 8; `L2` is stable in 3 of 8. Mean absolute z-scores: `SR=5.19`, `SR2=2.63`, `L1=3.96`, `L2=1.78`, `triple_var=4.37`. + +Poisson and prime-shuffle controls keep high `rank_all` while most observables are weak. At Poisson N=2048, `rank_all=1.952` but `stable_rank=1.036` and 4.62 of 5 observables are weak on average. This is the falsifying control for treating rank_all alone as a structural claim. + +## Findings + +1. **Perturbation rank is not interpretable without denominator gating.** In this perimeter, Poisson and prime-shuffle controls can show `rank_all` near 1.8-2.0. Because their original-vs-shuffle denominators are mostly weak, that rank is a retention-normalization artifact unless the stable-observable screen also supports it. + +2. **GUE does not show a stable second axis on canonical observables up to N=2048.** GUE `rank_all` falls from 1.913 at N=256 to 1.234 at N=2048; PC2 falls from 16.4% to 4.6%. After denominator gating, GUE stable rank stays close to 1.1-1.2. + +3. **The old L2-driven sign-flip should not be promoted without a denominator check.** Under canonical observables, GUE `L2` is weak relative to shuffle at every tested size and is stable in 0/8 replicates at N >= 512. This does not prove every L2 sign effect is false; it restricts such effects to local/sample-specific observations unless the denominator survives. + +4. **Primes become better conditioned with N, but not more multi-axis.** Prime windows gain stable observables as N grows: weak count drops from 4.75 at N=256 to 0.75 at N=2048. The effective rank does not grow with this conditioning; it is 1.442 at N=2048, and stable_rank is 1.462. + +## Verdict +**CONSTRAINT on META + BOUNDARY**: perturbation dimensionality must be reported as: + +> effective rank + PC2 + observable registry version + original-vs-shuffle z gate per observable. + +The cycle 03:30 "second GUE axis" remains restricted by cycle 06:25 and is further narrowed here: under canonical observables and the tested size curve, the stable statement is not "GUE has a second perturbation axis"; it is: + +> all-observable perturbation rank can inflate in weak-denominator regimes; after denominator gating, GUE and primes are both close to one perturbation coordinate in this perimeter, while Poisson/shuffle controls show why ungated rank is not structural evidence. + +## Consecutio +What opens now: the lab can keep using perturbation rank, but only as a gated observable. The next useful movement is not more PCA; it is an operator-level denominator map: for each observable, identify the perturbation/domain/scale region where `original - shuffle` is a real signal rather than a noisy divisor. + +## Auto-audit: 5 lenti +- **L1 hard constraint vs bias**: no zero/always claim. "Weak" means `abs(z) < 2` in the declared gate, not absence of signal. +- **L2 quantity vs ratio**: retention ratios are not read alone; raw denominator z-scores are reported first. +- **L3 no silent patching**: the 03:30 claim is explicitly restricted; it is not renamed as confirmed. +- **L4 edge cases**: short-GUE and low-N effects are isolated by size. The N=2048 perimeter is stated, not generalized. +- **L5 re-discovery**: PCA rank inflation from noisy normalization is a standard statistical risk. This is a lab constraint on method, not a new RMT result. + +## Files +- Script: `tools/exp_perturbation_rank_size_curve.py` +- Data: `tools/data/perturbation_rank_size_curve.json` +- Report: `tools/data/reports/agent_20260506_1941.md` diff --git a/tools/data/seme.json b/tools/data/seme.json index 9a30fb093a0fab963e44d5bed81d349bc1755ce9..d2b2b5f07ec443b64416cb27616b846fcb5403dd --- a/tools/data/seme.json +++ b/tools/data/seme.json @@ -98,6 +98,17 @@ "condensato_ref": "A14,A8", "origine": "cristallizzato 06/05 dalla consecutio del cycle 20260506_0625 (autopoietico self-finding)", "added_at": "2026-05-06T07:03:58.213606+00:00" + }, + { + "id": "PERTURBATION_DENOMINATOR_GATE", + "tipo": "vincolo", + "claim": "La dimensionalita di perturbazione va riportata solo insieme a PC2, versione observables_registry e gate original-vs-shuffle per osservabile. Nel perimetro 20260506_1941, Poisson e shuffle-primi producono rank_all ~1.8-2.0 con denominatori deboli; dopo gate abs(z)>=2 il rank stabile torna vicino a 1. Rank PCA non gated non e evidenza strutturale.", + "intensita": 0.95, + "porta": "META_BOUNDARY", + "manuale": true, + "condensato_ref": "A4,A8,A14,C2", + "origine": "cycle agent_20260506_1941: perturbation rank size curve canonical observables", + "added_at": "2026-05-06T19:41:00+00:00" } ], "potenziale_bloccato": [], @@ -276,4 +287,4 @@ "timestamp": "2026-04-04T19:34:08.014Z" } ] -} \ No newline at end of file +} diff --git a/tools/exp_perturbation_rank_size_curve.py b/tools/exp_perturbation_rank_size_curve.py new file mode 100644 index 0000000000000000000000000000000000000000..032847ca3d82ec3233ec2b41ad230dfe394d2791 --- /dev/null +++ b/tools/exp_perturbation_rank_size_curve.py @@ -0,0 +1,387 @@ +#!/usr/bin/env python3 +""" +exp_perturbation_rank_size_curve.py + +Reusable META audit for perturbation dimensionality. + +The 2026-05-06 06:25 cycle restricted the claim "GUE has a second +perturbation axis" to sample size, generator, and observable definitions. +This tool measures the size curve directly, using the canonical observable +registry and explicit original-vs-shuffle denominator diagnostics. + +The report owns interpretation. This script only measures: +- effective rank and PC2 across scale-selective perturbation profiles; +- original-vs-shuffle z-score per observable; +- whether apparent rank co-occurs with weak retention denominators. +""" + +from __future__ import annotations + +import argparse +import json +from pathlib import Path + +import numpy as np + +from observables_registry import ( + OBSERVABLES_CANONICAL, + OBSERVABLES_REGISTRY_VERSION, + compute_canonical, +) + + +OBS_NAMES = list(OBSERVABLES_CANONICAL.keys()) +PERT_NAMES = ["adjacent_swap", "block_shuffle", "large_gap_only", "uniform"] + + +def prime_gaps(n_gaps: int) -> np.ndarray: + """Return the first n_gaps prime gaps using a compact numpy sieve.""" + limit = max(100, int(n_gaps * (np.log(n_gaps + 10) + np.log(np.log(n_gaps + 10)) + 5))) + while True: + sieve = np.ones(limit + 1, dtype=bool) + sieve[:2] = False + for p in range(2, int(limit**0.5) + 1): + if sieve[p]: + sieve[p * p : limit + 1 : p] = False + primes = np.flatnonzero(sieve) + if len(primes) >= n_gaps + 1: + return np.diff(primes[: n_gaps + 1]).astype(float) + limit *= 2 + + +def gue_spacings(matrix_size: int, min_spacings: int, rng: np.random.Generator) -> np.ndarray: + """Generate unfolded GUE spacings by concatenating independent matrices.""" + parts = [] + edge = max(2, matrix_size // 10) + while sum(len(x) for x in parts) < min_spacings: + real = rng.standard_normal((matrix_size, matrix_size)) + imag = rng.standard_normal((matrix_size, matrix_size)) + h = real + 1j * imag + h = (h + h.conj().T) / (2.0 * np.sqrt(matrix_size)) + eigs = np.sort(np.linalg.eigvalsh(h).real) + bulk = eigs[edge:-edge] + gaps = np.diff(bulk) + mean = float(np.mean(gaps)) + if mean > 1e-15: + parts.append(gaps / mean) + return np.concatenate(parts)[:min_spacings].astype(float) + + +def perturb_adjacent_swap(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + idx = np.arange(0, len(out) - 1, 2) + chosen = idx[rng.random(len(idx)) < alpha] + tmp = out[chosen].copy() + out[chosen] = out[chosen + 1] + out[chosen + 1] = tmp + return out + + +def perturb_block_shuffle(gaps: np.ndarray, alpha: float, rng: np.random.Generator, block_size: int = 64) -> np.ndarray: + out = gaps.copy() + n_blocks = len(out) // block_size + if n_blocks <= 0: + return rng.permutation(out) if alpha > 0 else out + k = int(round(alpha * n_blocks)) + if k <= 0: + return out + for block in rng.choice(n_blocks, size=min(k, n_blocks), replace=False): + start = block * block_size + end = min(start + block_size, len(out)) + rng.shuffle(out[start:end]) + return out + + +def perturb_large_gap_only(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + idx = np.flatnonzero(out > np.median(out)) + k = int(round(alpha * len(idx))) + if k < 2: + return out + chosen = rng.choice(idx, size=min(k, len(idx)), replace=False) + vals = out[chosen].copy() + rng.shuffle(vals) + out[chosen] = vals + return out + + +def perturb_uniform(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + k = int(round(alpha * len(out))) + if k < 2: + return out + chosen = rng.choice(len(out), size=min(k, len(out)), replace=False) + vals = out[chosen].copy() + rng.shuffle(vals) + out[chosen] = vals + return out + + +PERTURB = { + "adjacent_swap": perturb_adjacent_swap, + "block_shuffle": perturb_block_shuffle, + "large_gap_only": perturb_large_gap_only, + "uniform": perturb_uniform, +} + + +def pca_summary(vectors: list[list[float]], labels: list[str]) -> dict: + matrix = np.array(vectors, dtype=float) + matrix = matrix - np.mean(matrix, axis=0, keepdims=True) + if min(matrix.shape) == 0 or np.max(np.abs(matrix)) <= 1e-15: + return { + "explained_variance": [], + "effective_rank": 0.0, + "centroid_cosine": {}, + "pc2": 0.0, + } + _, singular, _ = np.linalg.svd(matrix, full_matrices=False) + energy = singular * singular + if float(np.sum(energy)) <= 1e-15: + explained = np.zeros_like(energy) + effective_rank = 0.0 + else: + explained = energy / np.sum(energy) + pos = explained[explained > 1e-15] + effective_rank = float(np.exp(-np.sum(pos * np.log(pos)))) + + centroids = {} + for name in PERT_NAMES: + vals = np.array([v for v, label in zip(vectors, labels) if label == name], dtype=float) + if len(vals): + centroids[name] = np.mean(vals, axis=0) + + cosine = {} + for i, a_name in enumerate(PERT_NAMES): + for b_name in PERT_NAMES[i + 1 :]: + if a_name not in centroids or b_name not in centroids: + continue + a = centroids[a_name] + b = centroids[b_name] + denom = np.linalg.norm(a) * np.linalg.norm(b) + cosine[f"{a_name}_vs_{b_name}"] = float(np.dot(a, b) / denom) if denom > 1e-15 else 0.0 + + return { + "explained_variance": [float(x) for x in explained], + "effective_rank": effective_rank, + "centroid_cosine": cosine, + "pc2": float(explained[1]) if len(explained) > 1 else 0.0, + } + + +def analyze_sequence( + gaps: np.ndarray, + alphas: list[float], + n_trials: int, + n_baseline: int, + z_min: float, + rng: np.random.Generator, +) -> dict: + original = compute_canonical(gaps) + baseline_vals = {obs: [] for obs in OBS_NAMES} + for _ in range(n_baseline): + obs = compute_canonical(rng.permutation(gaps)) + for name in OBS_NAMES: + baseline_vals[name].append(obs[name]) + + baseline = {} + z = {} + denom = {} + for name in OBS_NAMES: + vals = np.array(baseline_vals[name], dtype=float) + mean = float(np.mean(vals)) + std = float(np.std(vals, ddof=1)) if len(vals) > 1 else 0.0 + baseline[name] = {"mean": mean, "std": std} + denom[name] = float(original[name] - mean) + z[name] = float(denom[name] / std) if std > 1e-15 else 0.0 + + stable_obs = [name for name in OBS_NAMES if abs(z[name]) >= z_min] + all_vectors = [] + screened_vectors = [] + labels = [] + profiles = [] + + for pert_name in PERT_NAMES: + for alpha in alphas: + trial_vals = {obs: [] for obs in OBS_NAMES} + for _ in range(n_trials): + perturbed = PERTURB[pert_name](gaps, alpha, rng) + obs = compute_canonical(perturbed) + for name in OBS_NAMES: + trial_vals[name].append(obs[name]) + means = {name: float(np.mean(trial_vals[name])) for name in OBS_NAMES} + retention = {} + for name in OBS_NAMES: + retention[name] = ( + float((means[name] - baseline[name]["mean"]) / denom[name]) + if abs(denom[name]) > 1e-12 + else 0.0 + ) + all_vector = [retention[name] for name in OBS_NAMES] + screened_vector = [retention[name] for name in stable_obs] + all_vectors.append(all_vector) + if len(stable_obs) >= 2: + screened_vectors.append(screened_vector) + labels.append(pert_name) + profiles.append( + { + "perturbation": pert_name, + "alpha": float(alpha), + "retention": retention, + "retention_vector": all_vector, + } + ) + + all_pca = pca_summary(all_vectors, labels) + screened_pca = pca_summary(screened_vectors, labels) if len(stable_obs) >= 2 else None + + return { + "n_gaps": int(len(gaps)), + "original": original, + "full_shuffle_baseline": baseline, + "denominator": denom, + "original_vs_shuffle_z": z, + "stable_observables": stable_obs, + "weak_observable_count": int(len(OBS_NAMES) - len(stable_obs)), + "profiles": profiles, + "pca_all_observables": all_pca, + "pca_stable_observables": screened_pca, + } + + +def summarize_replicates(items: list[dict]) -> dict: + def arr(path: tuple[str, ...]) -> np.ndarray: + vals = [] + for item in items: + x = item + for key in path: + if x is None: + break + x = x.get(key) + if isinstance(x, (int, float)): + vals.append(float(x)) + return np.array(vals, dtype=float) + + rank = arr(("pca_all_observables", "effective_rank")) + pc2 = arr(("pca_all_observables", "pc2")) + weak = np.array([item["weak_observable_count"] for item in items], dtype=float) + stable_rank = arr(("pca_stable_observables", "effective_rank")) + cos = arr(("pca_all_observables", "centroid_cosine", "adjacent_swap_vs_large_gap_only")) + + out = { + "n_replicates": len(items), + "rank_mean": float(np.mean(rank)) if len(rank) else 0.0, + "rank_std": float(np.std(rank, ddof=1)) if len(rank) > 1 else 0.0, + "pc2_mean": float(np.mean(pc2)) if len(pc2) else 0.0, + "pc2_std": float(np.std(pc2, ddof=1)) if len(pc2) > 1 else 0.0, + "weak_observable_count_mean": float(np.mean(weak)) if len(weak) else 0.0, + "stable_rank_mean": float(np.mean(stable_rank)) if len(stable_rank) else None, + "stable_rank_std": float(np.std(stable_rank, ddof=1)) if len(stable_rank) > 1 else 0.0, + "adjacent_vs_large_cosine_mean": float(np.mean(cos)) if len(cos) else 0.0, + "adjacent_vs_large_cosine_std": float(np.std(cos, ddof=1)) if len(cos) > 1 else 0.0, + } + if len(rank) > 1 and np.std(weak) > 1e-15 and np.std(rank) > 1e-15: + out["rank_vs_weak_count_corr"] = float(np.corrcoef(rank, weak)[0, 1]) + else: + out["rank_vs_weak_count_corr"] = 0.0 + return out + + +def build_prime_windows(max_n: int, n_reps: int) -> list[np.ndarray]: + total = max_n * n_reps + max_n + gaps = prime_gaps(total) + max_start = len(gaps) - max_n + starts = np.linspace(0, max_start, n_reps, dtype=int) + return [gaps[start : start + max_n].astype(float) for start in starts] + + +def run(args: argparse.Namespace) -> dict: + root_rng = np.random.default_rng(args.seed) + sizes = [int(x) for x in args.sizes.split(",") if x.strip()] + max_n = max(sizes) + alphas = [float(x) for x in np.linspace(args.alpha_min, args.alpha_max, args.n_alpha)] + + output = { + "experiment": "perturbation_rank_size_curve", + "question": "Does perturbation effective-rank stabilize with sample size under canonical observables?", + "observables_registry": OBSERVABLES_REGISTRY_VERSION, + "observables_used": OBS_NAMES, + "perturbations": PERT_NAMES, + "params": vars(args), + "alphas": alphas, + "domains": {}, + "summary": {}, + } + + prime_windows = build_prime_windows(max_n, args.n_replicates) + + print(f"observables_registry={OBSERVABLES_REGISTRY_VERSION}") + print(f"{'domain':<22} {'N':>6} {'rank':>7} {'PC2':>7} {'weak':>5} {'stable_rank':>11}") + + domain_builders = { + "primes_windows": lambda rep_rng, rep_i: prime_windows[rep_i], + "prime_shuffle_control": lambda rep_rng, rep_i: rep_rng.permutation(prime_windows[rep_i]), + "poisson": lambda rep_rng, rep_i: rep_rng.exponential(1.0, size=max_n), + "gue": lambda rep_rng, rep_i: gue_spacings(args.gue_matrix_size, max_n, rep_rng), + } + + for domain_name, builder in domain_builders.items(): + output["domains"][domain_name] = {} + output["summary"][domain_name] = {} + bases = [] + for rep_i in range(args.n_replicates): + rep_rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) + bases.append(builder(rep_rng, rep_i)) + + for n in sizes: + rows = [] + for rep_i, base in enumerate(bases): + rep_rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) + res = analyze_sequence( + base[:n], + alphas=alphas, + n_trials=args.n_trials, + n_baseline=args.n_baseline, + z_min=args.z_min, + rng=rep_rng, + ) + res["replicate"] = rep_i + rows.append(res) + summary = summarize_replicates(rows) + output["domains"][domain_name][str(n)] = rows + output["summary"][domain_name][str(n)] = summary + stable_rank = summary["stable_rank_mean"] + stable_text = f"{stable_rank:>11.3f}" if stable_rank is not None else f"{'NA':>11}" + print( + f"{domain_name:<22} {n:>6} {summary['rank_mean']:>7.3f} " + f"{summary['pc2_mean']:>7.3f} {summary['weak_observable_count_mean']:>5.2f} " + f"{stable_text}" + ) + + out_path = Path(args.out) + out_path.parent.mkdir(parents=True, exist_ok=True) + with out_path.open("w") as f: + json.dump(output, f, indent=2) + print(f"saved {out_path}") + return output + + +def main() -> None: + parser = argparse.ArgumentParser() + parser.add_argument("--sizes", default="128,256,512,1024,2048") + parser.add_argument("--n-replicates", type=int, default=8) + parser.add_argument("--gue-matrix-size", type=int, default=180) + parser.add_argument("--n-alpha", type=int, default=5) + parser.add_argument("--alpha-min", type=float, default=0.1) + parser.add_argument("--alpha-max", type=float, default=0.9) + parser.add_argument("--n-trials", type=int, default=8) + parser.add_argument("--n-baseline", type=int, default=16) + parser.add_argument("--z-min", type=float, default=2.0) + parser.add_argument("--seed", type=int, default=20260506) + parser.add_argument("--out", default="tools/data/perturbation_rank_size_curve.json") + args = parser.parse_args() + run(args) + + +if __name__ == "__main__": + main() succeeded in 151ms: exec /bin/bash -lc 'ln -sfn agent_20260506_1941.md tools/data/reports/latest.md && ls -l tools/data/reports/latest.md' in /opt/MM_D-ND succeeded in 0ms: lrwxrwxrwx 1 root root 22 May 6 19:45 tools/data/reports/latest.md -> agent_20260506_1941.md diff --git a/tools/data/reports/agent_20260506_1941.md b/tools/data/reports/agent_20260506_1941.md new file mode 100644 index 0000000000000000000000000000000000000000..5d74b086d2d0134f6b5a9dc93f036fd8b669c104 --- /dev/null +++ b/tools/data/reports/agent_20260506_1941.md @@ -0,0 +1,90 @@ +# Agent Report — Perturbation Rank Needs Denominator Gating +**Date**: 2026-05-06 19:41 +**Piano**: 66 +**Tension explored**: META + BOUNDARY +observables_registry: 1.0.0-2026-05-06 +observables_used: [SR, SR2, L1, L2, triple_var] + +## Claim Under Test +The valid next test from cycle 06:25 was a replicate-and-size curve for perturbation effective rank, with observable definitions versioned. This run asks: + +> Does the second perturbation axis stabilize as sample size grows, or is rank inflated when retention denominators are weak relative to full-shuffle baselines? + +## Experiment +Tool created: `tools/exp_perturbation_rank_size_curve.py` + +Atomic perimeter: +- domains: prime-gap windows, prime-shuffle controls, iid Poisson spacings, independent GUE spacings; +- sample sizes: 128, 256, 512, 1024, 2048 gaps; +- replicates/windows: 8 per domain-size point; +- perturbations: `adjacent_swap`, `block_shuffle`, `large_gap_only`, `uniform`; +- alpha grid: 0.1, 0.3, 0.5, 0.7, 0.9; +- trials per perturbation-alpha: 8; +- full-shuffle baselines: 16; +- canonical observables imported from `tools/observables_registry.py`; +- denominator gate: observable is stable only when `abs(original - shuffle_mean) / shuffle_std >= 2`. + +The script reports two ranks: +- `rank_all`: PCA effective rank using all five canonical observables; +- `stable_rank`: PCA effective rank after dropping observables whose original-vs-shuffle denominator is weak. + +## Results + +### Size Curve Summary + +| Domain | N | rank_all | PC2 | weak obs / 5 | stable_rank | +|---|---:|---:|---:|---:|---:| +| primes_windows | 128 | 1.789 ± 0.469 | 0.155 | 4.50 | 1.382 | +| primes_windows | 256 | 1.947 ± 0.645 | 0.174 | 4.75 | 1.262 | +| primes_windows | 512 | 1.892 ± 0.372 | 0.142 | 2.88 | 1.310 | +| primes_windows | 1024 | 1.679 ± 0.409 | 0.117 | 1.62 | 1.415 | +| primes_windows | 2048 | 1.442 ± 0.213 | 0.081 | 0.75 | 1.462 | +| prime_shuffle_control | 2048 | 1.797 ± 0.375 | 0.134 | 3.62 | 1.428 | +| poisson | 2048 | 1.952 ± 0.499 | 0.175 | 4.62 | 1.036 | +| gue | 128 | 1.703 ± 0.348 | 0.126 | 2.38 | 1.226 | +| gue | 256 | 1.913 ± 0.453 | 0.164 | 2.25 | 1.141 | +| gue | 512 | 1.542 ± 0.313 | 0.111 | 1.88 | 1.162 | +| gue | 1024 | 1.551 ± 0.395 | 0.105 | 1.88 | 1.157 | +| gue | 2048 | 1.234 ± 0.224 | 0.046 | 2.00 | 1.111 | + +### Observable Stability + +At GUE N=2048, `SR`, `L1`, and `triple_var` are stable in all 8 replicates; `SR2` and `L2` are stable in 0 of 8. Mean absolute z-scores: `SR=8.38`, `SR2=0.67`, `L1=11.58`, `L2=0.89`, `triple_var=11.66`. + +At primes N=2048, `SR`, `L1`, and `triple_var` are stable in all 8 windows; `SR2` is stable in 7 of 8; `L2` is stable in 3 of 8. Mean absolute z-scores: `SR=5.19`, `SR2=2.63`, `L1=3.96`, `L2=1.78`, `triple_var=4.37`. + +Poisson and prime-shuffle controls keep high `rank_all` while most observables are weak. At Poisson N=2048, `rank_all=1.952` but `stable_rank=1.036` and 4.62 of 5 observables are weak on average. This is the falsifying control for treating rank_all alone as a structural claim. + +## Findings + +1. **Perturbation rank is not interpretable without denominator gating.** In this perimeter, Poisson and prime-shuffle controls can show `rank_all` near 1.8-2.0. Because their original-vs-shuffle denominators are mostly weak, that rank is a retention-normalization artifact unless the stable-observable screen also supports it. + +2. **GUE does not show a stable second axis on canonical observables up to N=2048.** GUE `rank_all` falls from 1.913 at N=256 to 1.234 at N=2048; PC2 falls from 16.4% to 4.6%. After denominator gating, GUE stable rank stays close to 1.1-1.2. + +3. **The old L2-driven sign-flip should not be promoted without a denominator check.** Under canonical observables, GUE `L2` is weak relative to shuffle at every tested size and is stable in 0/8 replicates at N >= 512. This does not prove every L2 sign effect is false; it restricts such effects to local/sample-specific observations unless the denominator survives. + +4. **Primes become better conditioned with N, but not more multi-axis.** Prime windows gain stable observables as N grows: weak count drops from 4.75 at N=256 to 0.75 at N=2048. The effective rank does not grow with this conditioning; it is 1.442 at N=2048, and stable_rank is 1.462. + +## Verdict +**CONSTRAINT on META + BOUNDARY**: perturbation dimensionality must be reported as: + +> effective rank + PC2 + observable registry version + original-vs-shuffle z gate per observable. + +The cycle 03:30 "second GUE axis" remains restricted by cycle 06:25 and is further narrowed here: under canonical observables and the tested size curve, the stable statement is not "GUE has a second perturbation axis"; it is: + +> all-observable perturbation rank can inflate in weak-denominator regimes; after denominator gating, GUE and primes are both close to one perturbation coordinate in this perimeter, while Poisson/shuffle controls show why ungated rank is not structural evidence. + +## Consecutio +What opens now: the lab can keep using perturbation rank, but only as a gated observable. The next useful movement is not more PCA; it is an operator-level denominator map: for each observable, identify the perturbation/domain/scale region where `original - shuffle` is a real signal rather than a noisy divisor. + +## Auto-audit: 5 lenti +- **L1 hard constraint vs bias**: no zero/always claim. "Weak" means `abs(z) < 2` in the declared gate, not absence of signal. +- **L2 quantity vs ratio**: retention ratios are not read alone; raw denominator z-scores are reported first. +- **L3 no silent patching**: the 03:30 claim is explicitly restricted; it is not renamed as confirmed. +- **L4 edge cases**: short-GUE and low-N effects are isolated by size. The N=2048 perimeter is stated, not generalized. +- **L5 re-discovery**: PCA rank inflation from noisy normalization is a standard statistical risk. This is a lab constraint on method, not a new RMT result. + +## Files +- Script: `tools/exp_perturbation_rank_size_curve.py` +- Data: `tools/data/perturbation_rank_size_curve.json` +- Report: `tools/data/reports/agent_20260506_1941.md` diff --git a/tools/data/seme.json b/tools/data/seme.json index 9a30fb093a0fab963e44d5bed81d349bc1755ce9..d2b2b5f07ec443b64416cb27616b846fcb5403dd --- a/tools/data/seme.json +++ b/tools/data/seme.json @@ -98,6 +98,17 @@ "condensato_ref": "A14,A8", "origine": "cristallizzato 06/05 dalla consecutio del cycle 20260506_0625 (autopoietico self-finding)", "added_at": "2026-05-06T07:03:58.213606+00:00" + }, + { + "id": "PERTURBATION_DENOMINATOR_GATE", + "tipo": "vincolo", + "claim": "La dimensionalita di perturbazione va riportata solo insieme a PC2, versione observables_registry e gate original-vs-shuffle per osservabile. Nel perimetro 20260506_1941, Poisson e shuffle-primi producono rank_all ~1.8-2.0 con denominatori deboli; dopo gate abs(z)>=2 il rank stabile torna vicino a 1. Rank PCA non gated non e evidenza strutturale.", + "intensita": 0.95, + "porta": "META_BOUNDARY", + "manuale": true, + "condensato_ref": "A4,A8,A14,C2", + "origine": "cycle agent_20260506_1941: perturbation rank size curve canonical observables", + "added_at": "2026-05-06T19:41:00+00:00" } ], "potenziale_bloccato": [], @@ -276,4 +287,4 @@ "timestamp": "2026-04-04T19:34:08.014Z" } ] -} \ No newline at end of file +} diff --git a/tools/exp_perturbation_rank_size_curve.py b/tools/exp_perturbation_rank_size_curve.py new file mode 100644 index 0000000000000000000000000000000000000000..032847ca3d82ec3233ec2b41ad230dfe394d2791 --- /dev/null +++ b/tools/exp_perturbation_rank_size_curve.py @@ -0,0 +1,387 @@ +#!/usr/bin/env python3 +""" +exp_perturbation_rank_size_curve.py + +Reusable META audit for perturbation dimensionality. + +The 2026-05-06 06:25 cycle restricted the claim "GUE has a second +perturbation axis" to sample size, generator, and observable definitions. +This tool measures the size curve directly, using the canonical observable +registry and explicit original-vs-shuffle denominator diagnostics. + +The report owns interpretation. This script only measures: +- effective rank and PC2 across scale-selective perturbation profiles; +- original-vs-shuffle z-score per observable; +- whether apparent rank co-occurs with weak retention denominators. +""" + +from __future__ import annotations + +import argparse +import json +from pathlib import Path + +import numpy as np + +from observables_registry import ( + OBSERVABLES_CANONICAL, + OBSERVABLES_REGISTRY_VERSION, + compute_canonical, +) + + +OBS_NAMES = list(OBSERVABLES_CANONICAL.keys()) +PERT_NAMES = ["adjacent_swap", "block_shuffle", "large_gap_only", "uniform"] + + +def prime_gaps(n_gaps: int) -> np.ndarray: + """Return the first n_gaps prime gaps using a compact numpy sieve.""" + limit = max(100, int(n_gaps * (np.log(n_gaps + 10) + np.log(np.log(n_gaps + 10)) + 5))) + while True: + sieve = np.ones(limit + 1, dtype=bool) + sieve[:2] = False + for p in range(2, int(limit**0.5) + 1): + if sieve[p]: + sieve[p * p : limit + 1 : p] = False + primes = np.flatnonzero(sieve) + if len(primes) >= n_gaps + 1: + return np.diff(primes[: n_gaps + 1]).astype(float) + limit *= 2 + + +def gue_spacings(matrix_size: int, min_spacings: int, rng: np.random.Generator) -> np.ndarray: + """Generate unfolded GUE spacings by concatenating independent matrices.""" + parts = [] + edge = max(2, matrix_size // 10) + while sum(len(x) for x in parts) < min_spacings: + real = rng.standard_normal((matrix_size, matrix_size)) + imag = rng.standard_normal((matrix_size, matrix_size)) + h = real + 1j * imag + h = (h + h.conj().T) / (2.0 * np.sqrt(matrix_size)) + eigs = np.sort(np.linalg.eigvalsh(h).real) + bulk = eigs[edge:-edge] + gaps = np.diff(bulk) + mean = float(np.mean(gaps)) + if mean > 1e-15: + parts.append(gaps / mean) + return np.concatenate(parts)[:min_spacings].astype(float) + + +def perturb_adjacent_swap(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + idx = np.arange(0, len(out) - 1, 2) + chosen = idx[rng.random(len(idx)) < alpha] + tmp = out[chosen].copy() + out[chosen] = out[chosen + 1] + out[chosen + 1] = tmp + return out + + +def perturb_block_shuffle(gaps: np.ndarray, alpha: float, rng: np.random.Generator, block_size: int = 64) -> np.ndarray: + out = gaps.copy() + n_blocks = len(out) // block_size + if n_blocks <= 0: + return rng.permutation(out) if alpha > 0 else out + k = int(round(alpha * n_blocks)) + if k <= 0: + return out + for block in rng.choice(n_blocks, size=min(k, n_blocks), replace=False): + start = block * block_size + end = min(start + block_size, len(out)) + rng.shuffle(out[start:end]) + return out + + +def perturb_large_gap_only(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + idx = np.flatnonzero(out > np.median(out)) + k = int(round(alpha * len(idx))) + if k < 2: + return out + chosen = rng.choice(idx, size=min(k, len(idx)), replace=False) + vals = out[chosen].copy() + rng.shuffle(vals) + out[chosen] = vals + return out + + +def perturb_uniform(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + k = int(round(alpha * len(out))) + if k < 2: + return out + chosen = rng.choice(len(out), size=min(k, len(out)), replace=False) + vals = out[chosen].copy() + rng.shuffle(vals) + out[chosen] = vals + return out + + +PERTURB = { + "adjacent_swap": perturb_adjacent_swap, + "block_shuffle": perturb_block_shuffle, + "large_gap_only": perturb_large_gap_only, + "uniform": perturb_uniform, +} + + +def pca_summary(vectors: list[list[float]], labels: list[str]) -> dict: + matrix = np.array(vectors, dtype=float) + matrix = matrix - np.mean(matrix, axis=0, keepdims=True) + if min(matrix.shape) == 0 or np.max(np.abs(matrix)) <= 1e-15: + return { + "explained_variance": [], + "effective_rank": 0.0, + "centroid_cosine": {}, + "pc2": 0.0, + } + _, singular, _ = np.linalg.svd(matrix, full_matrices=False) + energy = singular * singular + if float(np.sum(energy)) <= 1e-15: + explained = np.zeros_like(energy) + effective_rank = 0.0 + else: + explained = energy / np.sum(energy) + pos = explained[explained > 1e-15] + effective_rank = float(np.exp(-np.sum(pos * np.log(pos)))) + + centroids = {} + for name in PERT_NAMES: + vals = np.array([v for v, label in zip(vectors, labels) if label == name], dtype=float) + if len(vals): + centroids[name] = np.mean(vals, axis=0) + + cosine = {} + for i, a_name in enumerate(PERT_NAMES): + for b_name in PERT_NAMES[i + 1 :]: + if a_name not in centroids or b_name not in centroids: + continue + a = centroids[a_name] + b = centroids[b_name] + denom = np.linalg.norm(a) * np.linalg.norm(b) + cosine[f"{a_name}_vs_{b_name}"] = float(np.dot(a, b) / denom) if denom > 1e-15 else 0.0 + + return { + "explained_variance": [float(x) for x in explained], + "effective_rank": effective_rank, + "centroid_cosine": cosine, + "pc2": float(explained[1]) if len(explained) > 1 else 0.0, + } + + +def analyze_sequence( + gaps: np.ndarray, + alphas: list[float], + n_trials: int, + n_baseline: int, + z_min: float, + rng: np.random.Generator, +) -> dict: + original = compute_canonical(gaps) + baseline_vals = {obs: [] for obs in OBS_NAMES} + for _ in range(n_baseline): + obs = compute_canonical(rng.permutation(gaps)) + for name in OBS_NAMES: + baseline_vals[name].append(obs[name]) + + baseline = {} + z = {} + denom = {} + for name in OBS_NAMES: + vals = np.array(baseline_vals[name], dtype=float) + mean = float(np.mean(vals)) + std = float(np.std(vals, ddof=1)) if len(vals) > 1 else 0.0 + baseline[name] = {"mean": mean, "std": std} + denom[name] = float(original[name] - mean) + z[name] = float(denom[name] / std) if std > 1e-15 else 0.0 + + stable_obs = [name for name in OBS_NAMES if abs(z[name]) >= z_min] + all_vectors = [] + screened_vectors = [] + labels = [] + profiles = [] + + for pert_name in PERT_NAMES: + for alpha in alphas: + trial_vals = {obs: [] for obs in OBS_NAMES} + for _ in range(n_trials): + perturbed = PERTURB[pert_name](gaps, alpha, rng) + obs = compute_canonical(perturbed) + for name in OBS_NAMES: + trial_vals[name].append(obs[name]) + means = {name: float(np.mean(trial_vals[name])) for name in OBS_NAMES} + retention = {} + for name in OBS_NAMES: + retention[name] = ( + float((means[name] - baseline[name]["mean"]) / denom[name]) + if abs(denom[name]) > 1e-12 + else 0.0 + ) + all_vector = [retention[name] for name in OBS_NAMES] + screened_vector = [retention[name] for name in stable_obs] + all_vectors.append(all_vector) + if len(stable_obs) >= 2: + screened_vectors.append(screened_vector) + labels.append(pert_name) + profiles.append( + { + "perturbation": pert_name, + "alpha": float(alpha), + "retention": retention, + "retention_vector": all_vector, + } + ) + + all_pca = pca_summary(all_vectors, labels) + screened_pca = pca_summary(screened_vectors, labels) if len(stable_obs) >= 2 else None + + return { + "n_gaps": int(len(gaps)), + "original": original, + "full_shuffle_baseline": baseline, + "denominator": denom, + "original_vs_shuffle_z": z, + "stable_observables": stable_obs, + "weak_observable_count": int(len(OBS_NAMES) - len(stable_obs)), + "profiles": profiles, + "pca_all_observables": all_pca, + "pca_stable_observables": screened_pca, + } + + +def summarize_replicates(items: list[dict]) -> dict: + def arr(path: tuple[str, ...]) -> np.ndarray: + vals = [] + for item in items: + x = item + for key in path: + if x is None: + break + x = x.get(key) + if isinstance(x, (int, float)): + vals.append(float(x)) + return np.array(vals, dtype=float) + + rank = arr(("pca_all_observables", "effective_rank")) + pc2 = arr(("pca_all_observables", "pc2")) + weak = np.array([item["weak_observable_count"] for item in items], dtype=float) + stable_rank = arr(("pca_stable_observables", "effective_rank")) + cos = arr(("pca_all_observables", "centroid_cosine", "adjacent_swap_vs_large_gap_only")) + + out = { + "n_replicates": len(items), + "rank_mean": float(np.mean(rank)) if len(rank) else 0.0, + "rank_std": float(np.std(rank, ddof=1)) if len(rank) > 1 else 0.0, + "pc2_mean": float(np.mean(pc2)) if len(pc2) else 0.0, + "pc2_std": float(np.std(pc2, ddof=1)) if len(pc2) > 1 else 0.0, + "weak_observable_count_mean": float(np.mean(weak)) if len(weak) else 0.0, + "stable_rank_mean": float(np.mean(stable_rank)) if len(stable_rank) else None, + "stable_rank_std": float(np.std(stable_rank, ddof=1)) if len(stable_rank) > 1 else 0.0, + "adjacent_vs_large_cosine_mean": float(np.mean(cos)) if len(cos) else 0.0, + "adjacent_vs_large_cosine_std": float(np.std(cos, ddof=1)) if len(cos) > 1 else 0.0, + } + if len(rank) > 1 and np.std(weak) > 1e-15 and np.std(rank) > 1e-15: + out["rank_vs_weak_count_corr"] = float(np.corrcoef(rank, weak)[0, 1]) + else: + out["rank_vs_weak_count_corr"] = 0.0 + return out + + +def build_prime_windows(max_n: int, n_reps: int) -> list[np.ndarray]: + total = max_n * n_reps + max_n + gaps = prime_gaps(total) + max_start = len(gaps) - max_n + starts = np.linspace(0, max_start, n_reps, dtype=int) + return [gaps[start : start + max_n].astype(float) for start in starts] + + +def run(args: argparse.Namespace) -> dict: + root_rng = np.random.default_rng(args.seed) + sizes = [int(x) for x in args.sizes.split(",") if x.strip()] + max_n = max(sizes) + alphas = [float(x) for x in np.linspace(args.alpha_min, args.alpha_max, args.n_alpha)] + + output = { + "experiment": "perturbation_rank_size_curve", + "question": "Does perturbation effective-rank stabilize with sample size under canonical observables?", + "observables_registry": OBSERVABLES_REGISTRY_VERSION, + "observables_used": OBS_NAMES, + "perturbations": PERT_NAMES, + "params": vars(args), + "alphas": alphas, + "domains": {}, + "summary": {}, + } + + prime_windows = build_prime_windows(max_n, args.n_replicates) + + print(f"observables_registry={OBSERVABLES_REGISTRY_VERSION}") + print(f"{'domain':<22} {'N':>6} {'rank':>7} {'PC2':>7} {'weak':>5} {'stable_rank':>11}") + + domain_builders = { + "primes_windows": lambda rep_rng, rep_i: prime_windows[rep_i], + "prime_shuffle_control": lambda rep_rng, rep_i: rep_rng.permutation(prime_windows[rep_i]), + "poisson": lambda rep_rng, rep_i: rep_rng.exponential(1.0, size=max_n), + "gue": lambda rep_rng, rep_i: gue_spacings(args.gue_matrix_size, max_n, rep_rng), + } + + for domain_name, builder in domain_builders.items(): + output["domains"][domain_name] = {} + output["summary"][domain_name] = {} + bases = [] + for rep_i in range(args.n_replicates): + rep_rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) + bases.append(builder(rep_rng, rep_i)) + + for n in sizes: + rows = [] + for rep_i, base in enumerate(bases): + rep_rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) + res = analyze_sequence( + base[:n], + alphas=alphas, + n_trials=args.n_trials, + n_baseline=args.n_baseline, + z_min=args.z_min, + rng=rep_rng, + ) + res["replicate"] = rep_i + rows.append(res) + summary = summarize_replicates(rows) + output["domains"][domain_name][str(n)] = rows + output["summary"][domain_name][str(n)] = summary + stable_rank = summary["stable_rank_mean"] + stable_text = f"{stable_rank:>11.3f}" if stable_rank is not None else f"{'NA':>11}" + print( + f"{domain_name:<22} {n:>6} {summary['rank_mean']:>7.3f} " + f"{summary['pc2_mean']:>7.3f} {summary['weak_observable_count_mean']:>5.2f} " + f"{stable_text}" + ) + + out_path = Path(args.out) + out_path.parent.mkdir(parents=True, exist_ok=True) + with out_path.open("w") as f: + json.dump(output, f, indent=2) + print(f"saved {out_path}") + return output + + +def main() -> None: + parser = argparse.ArgumentParser() + parser.add_argument("--sizes", default="128,256,512,1024,2048") + parser.add_argument("--n-replicates", type=int, default=8) + parser.add_argument("--gue-matrix-size", type=int, default=180) + parser.add_argument("--n-alpha", type=int, default=5) + parser.add_argument("--alpha-min", type=float, default=0.1) + parser.add_argument("--alpha-max", type=float, default=0.9) + parser.add_argument("--n-trials", type=int, default=8) + parser.add_argument("--n-baseline", type=int, default=16) + parser.add_argument("--z-min", type=float, default=2.0) + parser.add_argument("--seed", type=int, default=20260506) + parser.add_argument("--out", default="tools/data/perturbation_rank_size_curve.json") + args = parser.parse_args() + run(args) + + +if __name__ == "__main__": + main() exec /bin/bash -lc 'git status --short tools/exp_perturbation_rank_size_curve.py tools/data/perturbation_rank_size_curve.json tools/data/reports/agent_20260506_1941.md tools/data/reports/latest.md tools/data/seme.json' in /opt/MM_D-ND succeeded in 0ms: M tools/data/seme.json ?? tools/data/perturbation_rank_size_curve.json ?? tools/data/reports/agent_20260506_1941.md ?? tools/data/reports/latest.md ?? tools/exp_perturbation_rank_size_curve.py diff --git a/tools/data/reports/agent_20260506_1941.md b/tools/data/reports/agent_20260506_1941.md new file mode 100644 index 0000000000000000000000000000000000000000..5d74b086d2d0134f6b5a9dc93f036fd8b669c104 --- /dev/null +++ b/tools/data/reports/agent_20260506_1941.md @@ -0,0 +1,90 @@ +# Agent Report — Perturbation Rank Needs Denominator Gating +**Date**: 2026-05-06 19:41 +**Piano**: 66 +**Tension explored**: META + BOUNDARY +observables_registry: 1.0.0-2026-05-06 +observables_used: [SR, SR2, L1, L2, triple_var] + +## Claim Under Test +The valid next test from cycle 06:25 was a replicate-and-size curve for perturbation effective rank, with observable definitions versioned. This run asks: + +> Does the second perturbation axis stabilize as sample size grows, or is rank inflated when retention denominators are weak relative to full-shuffle baselines? + +## Experiment +Tool created: `tools/exp_perturbation_rank_size_curve.py` + +Atomic perimeter: +- domains: prime-gap windows, prime-shuffle controls, iid Poisson spacings, independent GUE spacings; +- sample sizes: 128, 256, 512, 1024, 2048 gaps; +- replicates/windows: 8 per domain-size point; +- perturbations: `adjacent_swap`, `block_shuffle`, `large_gap_only`, `uniform`; +- alpha grid: 0.1, 0.3, 0.5, 0.7, 0.9; +- trials per perturbation-alpha: 8; +- full-shuffle baselines: 16; +- canonical observables imported from `tools/observables_registry.py`; +- denominator gate: observable is stable only when `abs(original - shuffle_mean) / shuffle_std >= 2`. + +The script reports two ranks: +- `rank_all`: PCA effective rank using all five canonical observables; +- `stable_rank`: PCA effective rank after dropping observables whose original-vs-shuffle denominator is weak. + +## Results + +### Size Curve Summary + +| Domain | N | rank_all | PC2 | weak obs / 5 | stable_rank | +|---|---:|---:|---:|---:|---:| +| primes_windows | 128 | 1.789 ± 0.469 | 0.155 | 4.50 | 1.382 | +| primes_windows | 256 | 1.947 ± 0.645 | 0.174 | 4.75 | 1.262 | +| primes_windows | 512 | 1.892 ± 0.372 | 0.142 | 2.88 | 1.310 | +| primes_windows | 1024 | 1.679 ± 0.409 | 0.117 | 1.62 | 1.415 | +| primes_windows | 2048 | 1.442 ± 0.213 | 0.081 | 0.75 | 1.462 | +| prime_shuffle_control | 2048 | 1.797 ± 0.375 | 0.134 | 3.62 | 1.428 | +| poisson | 2048 | 1.952 ± 0.499 | 0.175 | 4.62 | 1.036 | +| gue | 128 | 1.703 ± 0.348 | 0.126 | 2.38 | 1.226 | +| gue | 256 | 1.913 ± 0.453 | 0.164 | 2.25 | 1.141 | +| gue | 512 | 1.542 ± 0.313 | 0.111 | 1.88 | 1.162 | +| gue | 1024 | 1.551 ± 0.395 | 0.105 | 1.88 | 1.157 | +| gue | 2048 | 1.234 ± 0.224 | 0.046 | 2.00 | 1.111 | + +### Observable Stability + +At GUE N=2048, `SR`, `L1`, and `triple_var` are stable in all 8 replicates; `SR2` and `L2` are stable in 0 of 8. Mean absolute z-scores: `SR=8.38`, `SR2=0.67`, `L1=11.58`, `L2=0.89`, `triple_var=11.66`. + +At primes N=2048, `SR`, `L1`, and `triple_var` are stable in all 8 windows; `SR2` is stable in 7 of 8; `L2` is stable in 3 of 8. Mean absolute z-scores: `SR=5.19`, `SR2=2.63`, `L1=3.96`, `L2=1.78`, `triple_var=4.37`. + +Poisson and prime-shuffle controls keep high `rank_all` while most observables are weak. At Poisson N=2048, `rank_all=1.952` but `stable_rank=1.036` and 4.62 of 5 observables are weak on average. This is the falsifying control for treating rank_all alone as a structural claim. + +## Findings + +1. **Perturbation rank is not interpretable without denominator gating.** In this perimeter, Poisson and prime-shuffle controls can show `rank_all` near 1.8-2.0. Because their original-vs-shuffle denominators are mostly weak, that rank is a retention-normalization artifact unless the stable-observable screen also supports it. + +2. **GUE does not show a stable second axis on canonical observables up to N=2048.** GUE `rank_all` falls from 1.913 at N=256 to 1.234 at N=2048; PC2 falls from 16.4% to 4.6%. After denominator gating, GUE stable rank stays close to 1.1-1.2. + +3. **The old L2-driven sign-flip should not be promoted without a denominator check.** Under canonical observables, GUE `L2` is weak relative to shuffle at every tested size and is stable in 0/8 replicates at N >= 512. This does not prove every L2 sign effect is false; it restricts such effects to local/sample-specific observations unless the denominator survives. + +4. **Primes become better conditioned with N, but not more multi-axis.** Prime windows gain stable observables as N grows: weak count drops from 4.75 at N=256 to 0.75 at N=2048. The effective rank does not grow with this conditioning; it is 1.442 at N=2048, and stable_rank is 1.462. + +## Verdict +**CONSTRAINT on META + BOUNDARY**: perturbation dimensionality must be reported as: + +> effective rank + PC2 + observable registry version + original-vs-shuffle z gate per observable. + +The cycle 03:30 "second GUE axis" remains restricted by cycle 06:25 and is further narrowed here: under canonical observables and the tested size curve, the stable statement is not "GUE has a second perturbation axis"; it is: + +> all-observable perturbation rank can inflate in weak-denominator regimes; after denominator gating, GUE and primes are both close to one perturbation coordinate in this perimeter, while Poisson/shuffle controls show why ungated rank is not structural evidence. + +## Consecutio +What opens now: the lab can keep using perturbation rank, but only as a gated observable. The next useful movement is not more PCA; it is an operator-level denominator map: for each observable, identify the perturbation/domain/scale region where `original - shuffle` is a real signal rather than a noisy divisor. + +## Auto-audit: 5 lenti +- **L1 hard constraint vs bias**: no zero/always claim. "Weak" means `abs(z) < 2` in the declared gate, not absence of signal. +- **L2 quantity vs ratio**: retention ratios are not read alone; raw denominator z-scores are reported first. +- **L3 no silent patching**: the 03:30 claim is explicitly restricted; it is not renamed as confirmed. +- **L4 edge cases**: short-GUE and low-N effects are isolated by size. The N=2048 perimeter is stated, not generalized. +- **L5 re-discovery**: PCA rank inflation from noisy normalization is a standard statistical risk. This is a lab constraint on method, not a new RMT result. + +## Files +- Script: `tools/exp_perturbation_rank_size_curve.py` +- Data: `tools/data/perturbation_rank_size_curve.json` +- Report: `tools/data/reports/agent_20260506_1941.md` diff --git a/tools/data/seme.json b/tools/data/seme.json index 9a30fb093a0fab963e44d5bed81d349bc1755ce9..d2b2b5f07ec443b64416cb27616b846fcb5403dd --- a/tools/data/seme.json +++ b/tools/data/seme.json @@ -98,6 +98,17 @@ "condensato_ref": "A14,A8", "origine": "cristallizzato 06/05 dalla consecutio del cycle 20260506_0625 (autopoietico self-finding)", "added_at": "2026-05-06T07:03:58.213606+00:00" + }, + { + "id": "PERTURBATION_DENOMINATOR_GATE", + "tipo": "vincolo", + "claim": "La dimensionalita di perturbazione va riportata solo insieme a PC2, versione observables_registry e gate original-vs-shuffle per osservabile. Nel perimetro 20260506_1941, Poisson e shuffle-primi producono rank_all ~1.8-2.0 con denominatori deboli; dopo gate abs(z)>=2 il rank stabile torna vicino a 1. Rank PCA non gated non e evidenza strutturale.", + "intensita": 0.95, + "porta": "META_BOUNDARY", + "manuale": true, + "condensato_ref": "A4,A8,A14,C2", + "origine": "cycle agent_20260506_1941: perturbation rank size curve canonical observables", + "added_at": "2026-05-06T19:41:00+00:00" } ], "potenziale_bloccato": [], @@ -276,4 +287,4 @@ "timestamp": "2026-04-04T19:34:08.014Z" } ] -} \ No newline at end of file +} diff --git a/tools/exp_perturbation_rank_size_curve.py b/tools/exp_perturbation_rank_size_curve.py new file mode 100644 index 0000000000000000000000000000000000000000..032847ca3d82ec3233ec2b41ad230dfe394d2791 --- /dev/null +++ b/tools/exp_perturbation_rank_size_curve.py @@ -0,0 +1,387 @@ +#!/usr/bin/env python3 +""" +exp_perturbation_rank_size_curve.py + +Reusable META audit for perturbation dimensionality. + +The 2026-05-06 06:25 cycle restricted the claim "GUE has a second +perturbation axis" to sample size, generator, and observable definitions. +This tool measures the size curve directly, using the canonical observable +registry and explicit original-vs-shuffle denominator diagnostics. + +The report owns interpretation. This script only measures: +- effective rank and PC2 across scale-selective perturbation profiles; +- original-vs-shuffle z-score per observable; +- whether apparent rank co-occurs with weak retention denominators. +""" + +from __future__ import annotations + +import argparse +import json +from pathlib import Path + +import numpy as np + +from observables_registry import ( + OBSERVABLES_CANONICAL, + OBSERVABLES_REGISTRY_VERSION, + compute_canonical, +) + + +OBS_NAMES = list(OBSERVABLES_CANONICAL.keys()) +PERT_NAMES = ["adjacent_swap", "block_shuffle", "large_gap_only", "uniform"] + + +def prime_gaps(n_gaps: int) -> np.ndarray: + """Return the first n_gaps prime gaps using a compact numpy sieve.""" + limit = max(100, int(n_gaps * (np.log(n_gaps + 10) + np.log(np.log(n_gaps + 10)) + 5))) + while True: + sieve = np.ones(limit + 1, dtype=bool) + sieve[:2] = False + for p in range(2, int(limit**0.5) + 1): + if sieve[p]: + sieve[p * p : limit + 1 : p] = False + primes = np.flatnonzero(sieve) + if len(primes) >= n_gaps + 1: + return np.diff(primes[: n_gaps + 1]).astype(float) + limit *= 2 + + +def gue_spacings(matrix_size: int, min_spacings: int, rng: np.random.Generator) -> np.ndarray: + """Generate unfolded GUE spacings by concatenating independent matrices.""" + parts = [] + edge = max(2, matrix_size // 10) + while sum(len(x) for x in parts) < min_spacings: + real = rng.standard_normal((matrix_size, matrix_size)) + imag = rng.standard_normal((matrix_size, matrix_size)) + h = real + 1j * imag + h = (h + h.conj().T) / (2.0 * np.sqrt(matrix_size)) + eigs = np.sort(np.linalg.eigvalsh(h).real) + bulk = eigs[edge:-edge] + gaps = np.diff(bulk) + mean = float(np.mean(gaps)) + if mean > 1e-15: + parts.append(gaps / mean) + return np.concatenate(parts)[:min_spacings].astype(float) + + +def perturb_adjacent_swap(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + idx = np.arange(0, len(out) - 1, 2) + chosen = idx[rng.random(len(idx)) < alpha] + tmp = out[chosen].copy() + out[chosen] = out[chosen + 1] + out[chosen + 1] = tmp + return out + + +def perturb_block_shuffle(gaps: np.ndarray, alpha: float, rng: np.random.Generator, block_size: int = 64) -> np.ndarray: + out = gaps.copy() + n_blocks = len(out) // block_size + if n_blocks <= 0: + return rng.permutation(out) if alpha > 0 else out + k = int(round(alpha * n_blocks)) + if k <= 0: + return out + for block in rng.choice(n_blocks, size=min(k, n_blocks), replace=False): + start = block * block_size + end = min(start + block_size, len(out)) + rng.shuffle(out[start:end]) + return out + + +def perturb_large_gap_only(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + idx = np.flatnonzero(out > np.median(out)) + k = int(round(alpha * len(idx))) + if k < 2: + return out + chosen = rng.choice(idx, size=min(k, len(idx)), replace=False) + vals = out[chosen].copy() + rng.shuffle(vals) + out[chosen] = vals + return out + + +def perturb_uniform(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + k = int(round(alpha * len(out))) + if k < 2: + return out + chosen = rng.choice(len(out), size=min(k, len(out)), replace=False) + vals = out[chosen].copy() + rng.shuffle(vals) + out[chosen] = vals + return out + + +PERTURB = { + "adjacent_swap": perturb_adjacent_swap, + "block_shuffle": perturb_block_shuffle, + "large_gap_only": perturb_large_gap_only, + "uniform": perturb_uniform, +} + + +def pca_summary(vectors: list[list[float]], labels: list[str]) -> dict: + matrix = np.array(vectors, dtype=float) + matrix = matrix - np.mean(matrix, axis=0, keepdims=True) + if min(matrix.shape) == 0 or np.max(np.abs(matrix)) <= 1e-15: + return { + "explained_variance": [], + "effective_rank": 0.0, + "centroid_cosine": {}, + "pc2": 0.0, + } + _, singular, _ = np.linalg.svd(matrix, full_matrices=False) + energy = singular * singular + if float(np.sum(energy)) <= 1e-15: + explained = np.zeros_like(energy) + effective_rank = 0.0 + else: + explained = energy / np.sum(energy) + pos = explained[explained > 1e-15] + effective_rank = float(np.exp(-np.sum(pos * np.log(pos)))) + + centroids = {} + for name in PERT_NAMES: + vals = np.array([v for v, label in zip(vectors, labels) if label == name], dtype=float) + if len(vals): + centroids[name] = np.mean(vals, axis=0) + + cosine = {} + for i, a_name in enumerate(PERT_NAMES): + for b_name in PERT_NAMES[i + 1 :]: + if a_name not in centroids or b_name not in centroids: + continue + a = centroids[a_name] + b = centroids[b_name] + denom = np.linalg.norm(a) * np.linalg.norm(b) + cosine[f"{a_name}_vs_{b_name}"] = float(np.dot(a, b) / denom) if denom > 1e-15 else 0.0 + + return { + "explained_variance": [float(x) for x in explained], + "effective_rank": effective_rank, + "centroid_cosine": cosine, + "pc2": float(explained[1]) if len(explained) > 1 else 0.0, + } + + +def analyze_sequence( + gaps: np.ndarray, + alphas: list[float], + n_trials: int, + n_baseline: int, + z_min: float, + rng: np.random.Generator, +) -> dict: + original = compute_canonical(gaps) + baseline_vals = {obs: [] for obs in OBS_NAMES} + for _ in range(n_baseline): + obs = compute_canonical(rng.permutation(gaps)) + for name in OBS_NAMES: + baseline_vals[name].append(obs[name]) + + baseline = {} + z = {} + denom = {} + for name in OBS_NAMES: + vals = np.array(baseline_vals[name], dtype=float) + mean = float(np.mean(vals)) + std = float(np.std(vals, ddof=1)) if len(vals) > 1 else 0.0 + baseline[name] = {"mean": mean, "std": std} + denom[name] = float(original[name] - mean) + z[name] = float(denom[name] / std) if std > 1e-15 else 0.0 + + stable_obs = [name for name in OBS_NAMES if abs(z[name]) >= z_min] + all_vectors = [] + screened_vectors = [] + labels = [] + profiles = [] + + for pert_name in PERT_NAMES: + for alpha in alphas: + trial_vals = {obs: [] for obs in OBS_NAMES} + for _ in range(n_trials): + perturbed = PERTURB[pert_name](gaps, alpha, rng) + obs = compute_canonical(perturbed) + for name in OBS_NAMES: + trial_vals[name].append(obs[name]) + means = {name: float(np.mean(trial_vals[name])) for name in OBS_NAMES} + retention = {} + for name in OBS_NAMES: + retention[name] = ( + float((means[name] - baseline[name]["mean"]) / denom[name]) + if abs(denom[name]) > 1e-12 + else 0.0 + ) + all_vector = [retention[name] for name in OBS_NAMES] + screened_vector = [retention[name] for name in stable_obs] + all_vectors.append(all_vector) + if len(stable_obs) >= 2: + screened_vectors.append(screened_vector) + labels.append(pert_name) + profiles.append( + { + "perturbation": pert_name, + "alpha": float(alpha), + "retention": retention, + "retention_vector": all_vector, + } + ) + + all_pca = pca_summary(all_vectors, labels) + screened_pca = pca_summary(screened_vectors, labels) if len(stable_obs) >= 2 else None + + return { + "n_gaps": int(len(gaps)), + "original": original, + "full_shuffle_baseline": baseline, + "denominator": denom, + "original_vs_shuffle_z": z, + "stable_observables": stable_obs, + "weak_observable_count": int(len(OBS_NAMES) - len(stable_obs)), + "profiles": profiles, + "pca_all_observables": all_pca, + "pca_stable_observables": screened_pca, + } + + +def summarize_replicates(items: list[dict]) -> dict: + def arr(path: tuple[str, ...]) -> np.ndarray: + vals = [] + for item in items: + x = item + for key in path: + if x is None: + break + x = x.get(key) + if isinstance(x, (int, float)): + vals.append(float(x)) + return np.array(vals, dtype=float) + + rank = arr(("pca_all_observables", "effective_rank")) + pc2 = arr(("pca_all_observables", "pc2")) + weak = np.array([item["weak_observable_count"] for item in items], dtype=float) + stable_rank = arr(("pca_stable_observables", "effective_rank")) + cos = arr(("pca_all_observables", "centroid_cosine", "adjacent_swap_vs_large_gap_only")) + + out = { + "n_replicates": len(items), + "rank_mean": float(np.mean(rank)) if len(rank) else 0.0, + "rank_std": float(np.std(rank, ddof=1)) if len(rank) > 1 else 0.0, + "pc2_mean": float(np.mean(pc2)) if len(pc2) else 0.0, + "pc2_std": float(np.std(pc2, ddof=1)) if len(pc2) > 1 else 0.0, + "weak_observable_count_mean": float(np.mean(weak)) if len(weak) else 0.0, + "stable_rank_mean": float(np.mean(stable_rank)) if len(stable_rank) else None, + "stable_rank_std": float(np.std(stable_rank, ddof=1)) if len(stable_rank) > 1 else 0.0, + "adjacent_vs_large_cosine_mean": float(np.mean(cos)) if len(cos) else 0.0, + "adjacent_vs_large_cosine_std": float(np.std(cos, ddof=1)) if len(cos) > 1 else 0.0, + } + if len(rank) > 1 and np.std(weak) > 1e-15 and np.std(rank) > 1e-15: + out["rank_vs_weak_count_corr"] = float(np.corrcoef(rank, weak)[0, 1]) + else: + out["rank_vs_weak_count_corr"] = 0.0 + return out + + +def build_prime_windows(max_n: int, n_reps: int) -> list[np.ndarray]: + total = max_n * n_reps + max_n + gaps = prime_gaps(total) + max_start = len(gaps) - max_n + starts = np.linspace(0, max_start, n_reps, dtype=int) + return [gaps[start : start + max_n].astype(float) for start in starts] + + +def run(args: argparse.Namespace) -> dict: + root_rng = np.random.default_rng(args.seed) + sizes = [int(x) for x in args.sizes.split(",") if x.strip()] + max_n = max(sizes) + alphas = [float(x) for x in np.linspace(args.alpha_min, args.alpha_max, args.n_alpha)] + + output = { + "experiment": "perturbation_rank_size_curve", + "question": "Does perturbation effective-rank stabilize with sample size under canonical observables?", + "observables_registry": OBSERVABLES_REGISTRY_VERSION, + "observables_used": OBS_NAMES, + "perturbations": PERT_NAMES, + "params": vars(args), + "alphas": alphas, + "domains": {}, + "summary": {}, + } + + prime_windows = build_prime_windows(max_n, args.n_replicates) + + print(f"observables_registry={OBSERVABLES_REGISTRY_VERSION}") + print(f"{'domain':<22} {'N':>6} {'rank':>7} {'PC2':>7} {'weak':>5} {'stable_rank':>11}") + + domain_builders = { + "primes_windows": lambda rep_rng, rep_i: prime_windows[rep_i], + "prime_shuffle_control": lambda rep_rng, rep_i: rep_rng.permutation(prime_windows[rep_i]), + "poisson": lambda rep_rng, rep_i: rep_rng.exponential(1.0, size=max_n), + "gue": lambda rep_rng, rep_i: gue_spacings(args.gue_matrix_size, max_n, rep_rng), + } + + for domain_name, builder in domain_builders.items(): + output["domains"][domain_name] = {} + output["summary"][domain_name] = {} + bases = [] + for rep_i in range(args.n_replicates): + rep_rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) + bases.append(builder(rep_rng, rep_i)) + + for n in sizes: + rows = [] + for rep_i, base in enumerate(bases): + rep_rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) + res = analyze_sequence( + base[:n], + alphas=alphas, + n_trials=args.n_trials, + n_baseline=args.n_baseline, + z_min=args.z_min, + rng=rep_rng, + ) + res["replicate"] = rep_i + rows.append(res) + summary = summarize_replicates(rows) + output["domains"][domain_name][str(n)] = rows + output["summary"][domain_name][str(n)] = summary + stable_rank = summary["stable_rank_mean"] + stable_text = f"{stable_rank:>11.3f}" if stable_rank is not None else f"{'NA':>11}" + print( + f"{domain_name:<22} {n:>6} {summary['rank_mean']:>7.3f} " + f"{summary['pc2_mean']:>7.3f} {summary['weak_observable_count_mean']:>5.2f} " + f"{stable_text}" + ) + + out_path = Path(args.out) + out_path.parent.mkdir(parents=True, exist_ok=True) + with out_path.open("w") as f: + json.dump(output, f, indent=2) + print(f"saved {out_path}") + return output + + +def main() -> None: + parser = argparse.ArgumentParser() + parser.add_argument("--sizes", default="128,256,512,1024,2048") + parser.add_argument("--n-replicates", type=int, default=8) + parser.add_argument("--gue-matrix-size", type=int, default=180) + parser.add_argument("--n-alpha", type=int, default=5) + parser.add_argument("--alpha-min", type=float, default=0.1) + parser.add_argument("--alpha-max", type=float, default=0.9) + parser.add_argument("--n-trials", type=int, default=8) + parser.add_argument("--n-baseline", type=int, default=16) + parser.add_argument("--z-min", type=float, default=2.0) + parser.add_argument("--seed", type=int, default=20260506) + parser.add_argument("--out", default="tools/data/perturbation_rank_size_curve.json") + args = parser.parse_args() + run(args) + + +if __name__ == "__main__": + main() codex Completato. Ho eseguito l’esperimento META/BOUNDARY sulla curva rank-vs-N con osservabili canonici e gate sui denominatori original-vs-shuffle. File prodotti: - Report: [agent_20260506_1941.md](/opt/MM_D-ND/tools/data/reports/agent_20260506_1941.md) - Script riusabile: [exp_perturbation_rank_size_curve.py](/opt/MM_D-ND/tools/exp_perturbation_rank_size_curve.py) - Dati: [perturbation_rank_size_curve.json](/opt/MM_D-ND/tools/data/perturbation_rank_size_curve.json) - Seme aggiornato: [seme.json](/opt/MM_D-ND/tools/data/seme.json) Risultante: il rank di perturbazione non è interpretabile senza gate sui denominatori. Poisson e shuffle-primi possono mostrare `rank_all ~1.8-2.0` con osservabili deboli; dopo gate `abs(z)>=2`, il rank torna vicino a 1. Nel perimetro testato, GUE non mostra un secondo asse stabile con osservabili canonici fino a N=2048. Verifiche eseguite: - `python -m py_compile tools/exp_perturbation_rank_size_curve.py` - validazione JSON di `seme.json` e `perturbation_rank_size_curve.json` - aggiornato `tools/data/reports/latest.md` al nuovo report. diff --git a/tools/data/reports/agent_20260506_1941.md b/tools/data/reports/agent_20260506_1941.md new file mode 100644 index 0000000000000000000000000000000000000000..5d74b086d2d0134f6b5a9dc93f036fd8b669c104 --- /dev/null +++ b/tools/data/reports/agent_20260506_1941.md @@ -0,0 +1,90 @@ +# Agent Report — Perturbation Rank Needs Denominator Gating +**Date**: 2026-05-06 19:41 +**Piano**: 66 +**Tension explored**: META + BOUNDARY +observables_registry: 1.0.0-2026-05-06 +observables_used: [SR, SR2, L1, L2, triple_var] + +## Claim Under Test +The valid next test from cycle 06:25 was a replicate-and-size curve for perturbation effective rank, with observable definitions versioned. This run asks: + +> Does the second perturbation axis stabilize as sample size grows, or is rank inflated when retention denominators are weak relative to full-shuffle baselines? + +## Experiment +Tool created: `tools/exp_perturbation_rank_size_curve.py` + +Atomic perimeter: +- domains: prime-gap windows, prime-shuffle controls, iid Poisson spacings, independent GUE spacings; +- sample sizes: 128, 256, 512, 1024, 2048 gaps; +- replicates/windows: 8 per domain-size point; +- perturbations: `adjacent_swap`, `block_shuffle`, `large_gap_only`, `uniform`; +- alpha grid: 0.1, 0.3, 0.5, 0.7, 0.9; +- trials per perturbation-alpha: 8; +- full-shuffle baselines: 16; +- canonical observables imported from `tools/observables_registry.py`; +- denominator gate: observable is stable only when `abs(original - shuffle_mean) / shuffle_std >= 2`. + +The script reports two ranks: +- `rank_all`: PCA effective rank using all five canonical observables; +- `stable_rank`: PCA effective rank after dropping observables whose original-vs-shuffle denominator is weak. + +## Results + +### Size Curve Summary + +| Domain | N | rank_all | PC2 | weak obs / 5 | stable_rank | +|---|---:|---:|---:|---:|---:| +| primes_windows | 128 | 1.789 ± 0.469 | 0.155 | 4.50 | 1.382 | +| primes_windows | 256 | 1.947 ± 0.645 | 0.174 | 4.75 | 1.262 | +| primes_windows | 512 | 1.892 ± 0.372 | 0.142 | 2.88 | 1.310 | +| primes_windows | 1024 | 1.679 ± 0.409 | 0.117 | 1.62 | 1.415 | +| primes_windows | 2048 | 1.442 ± 0.213 | 0.081 | 0.75 | 1.462 | +| prime_shuffle_control | 2048 | 1.797 ± 0.375 | 0.134 | 3.62 | 1.428 | +| poisson | 2048 | 1.952 ± 0.499 | 0.175 | 4.62 | 1.036 | +| gue | 128 | 1.703 ± 0.348 | 0.126 | 2.38 | 1.226 | +| gue | 256 | 1.913 ± 0.453 | 0.164 | 2.25 | 1.141 | +| gue | 512 | 1.542 ± 0.313 | 0.111 | 1.88 | 1.162 | +| gue | 1024 | 1.551 ± 0.395 | 0.105 | 1.88 | 1.157 | +| gue | 2048 | 1.234 ± 0.224 | 0.046 | 2.00 | 1.111 | + +### Observable Stability + +At GUE N=2048, `SR`, `L1`, and `triple_var` are stable in all 8 replicates; `SR2` and `L2` are stable in 0 of 8. Mean absolute z-scores: `SR=8.38`, `SR2=0.67`, `L1=11.58`, `L2=0.89`, `triple_var=11.66`. + +At primes N=2048, `SR`, `L1`, and `triple_var` are stable in all 8 windows; `SR2` is stable in 7 of 8; `L2` is stable in 3 of 8. Mean absolute z-scores: `SR=5.19`, `SR2=2.63`, `L1=3.96`, `L2=1.78`, `triple_var=4.37`. + +Poisson and prime-shuffle controls keep high `rank_all` while most observables are weak. At Poisson N=2048, `rank_all=1.952` but `stable_rank=1.036` and 4.62 of 5 observables are weak on average. This is the falsifying control for treating rank_all alone as a structural claim. + +## Findings + +1. **Perturbation rank is not interpretable without denominator gating.** In this perimeter, Poisson and prime-shuffle controls can show `rank_all` near 1.8-2.0. Because their original-vs-shuffle denominators are mostly weak, that rank is a retention-normalization artifact unless the stable-observable screen also supports it. + +2. **GUE does not show a stable second axis on canonical observables up to N=2048.** GUE `rank_all` falls from 1.913 at N=256 to 1.234 at N=2048; PC2 falls from 16.4% to 4.6%. After denominator gating, GUE stable rank stays close to 1.1-1.2. + +3. **The old L2-driven sign-flip should not be promoted without a denominator check.** Under canonical observables, GUE `L2` is weak relative to shuffle at every tested size and is stable in 0/8 replicates at N >= 512. This does not prove every L2 sign effect is false; it restricts such effects to local/sample-specific observations unless the denominator survives. + +4. **Primes become better conditioned with N, but not more multi-axis.** Prime windows gain stable observables as N grows: weak count drops from 4.75 at N=256 to 0.75 at N=2048. The effective rank does not grow with this conditioning; it is 1.442 at N=2048, and stable_rank is 1.462. + +## Verdict +**CONSTRAINT on META + BOUNDARY**: perturbation dimensionality must be reported as: + +> effective rank + PC2 + observable registry version + original-vs-shuffle z gate per observable. + +The cycle 03:30 "second GUE axis" remains restricted by cycle 06:25 and is further narrowed here: under canonical observables and the tested size curve, the stable statement is not "GUE has a second perturbation axis"; it is: + +> all-observable perturbation rank can inflate in weak-denominator regimes; after denominator gating, GUE and primes are both close to one perturbation coordinate in this perimeter, while Poisson/shuffle controls show why ungated rank is not structural evidence. + +## Consecutio +What opens now: the lab can keep using perturbation rank, but only as a gated observable. The next useful movement is not more PCA; it is an operator-level denominator map: for each observable, identify the perturbation/domain/scale region where `original - shuffle` is a real signal rather than a noisy divisor. + +## Auto-audit: 5 lenti +- **L1 hard constraint vs bias**: no zero/always claim. "Weak" means `abs(z) < 2` in the declared gate, not absence of signal. +- **L2 quantity vs ratio**: retention ratios are not read alone; raw denominator z-scores are reported first. +- **L3 no silent patching**: the 03:30 claim is explicitly restricted; it is not renamed as confirmed. +- **L4 edge cases**: short-GUE and low-N effects are isolated by size. The N=2048 perimeter is stated, not generalized. +- **L5 re-discovery**: PCA rank inflation from noisy normalization is a standard statistical risk. This is a lab constraint on method, not a new RMT result. + +## Files +- Script: `tools/exp_perturbation_rank_size_curve.py` +- Data: `tools/data/perturbation_rank_size_curve.json` +- Report: `tools/data/reports/agent_20260506_1941.md` diff --git a/tools/data/seme.json b/tools/data/seme.json index 9a30fb093a0fab963e44d5bed81d349bc1755ce9..d2b2b5f07ec443b64416cb27616b846fcb5403dd --- a/tools/data/seme.json +++ b/tools/data/seme.json @@ -98,6 +98,17 @@ "condensato_ref": "A14,A8", "origine": "cristallizzato 06/05 dalla consecutio del cycle 20260506_0625 (autopoietico self-finding)", "added_at": "2026-05-06T07:03:58.213606+00:00" + }, + { + "id": "PERTURBATION_DENOMINATOR_GATE", + "tipo": "vincolo", + "claim": "La dimensionalita di perturbazione va riportata solo insieme a PC2, versione observables_registry e gate original-vs-shuffle per osservabile. Nel perimetro 20260506_1941, Poisson e shuffle-primi producono rank_all ~1.8-2.0 con denominatori deboli; dopo gate abs(z)>=2 il rank stabile torna vicino a 1. Rank PCA non gated non e evidenza strutturale.", + "intensita": 0.95, + "porta": "META_BOUNDARY", + "manuale": true, + "condensato_ref": "A4,A8,A14,C2", + "origine": "cycle agent_20260506_1941: perturbation rank size curve canonical observables", + "added_at": "2026-05-06T19:41:00+00:00" } ], "potenziale_bloccato": [], @@ -276,4 +287,4 @@ "timestamp": "2026-04-04T19:34:08.014Z" } ] -} \ No newline at end of file +} diff --git a/tools/exp_perturbation_rank_size_curve.py b/tools/exp_perturbation_rank_size_curve.py new file mode 100644 index 0000000000000000000000000000000000000000..032847ca3d82ec3233ec2b41ad230dfe394d2791 --- /dev/null +++ b/tools/exp_perturbation_rank_size_curve.py @@ -0,0 +1,387 @@ +#!/usr/bin/env python3 +""" +exp_perturbation_rank_size_curve.py + +Reusable META audit for perturbation dimensionality. + +The 2026-05-06 06:25 cycle restricted the claim "GUE has a second +perturbation axis" to sample size, generator, and observable definitions. +This tool measures the size curve directly, using the canonical observable +registry and explicit original-vs-shuffle denominator diagnostics. + +The report owns interpretation. This script only measures: +- effective rank and PC2 across scale-selective perturbation profiles; +- original-vs-shuffle z-score per observable; +- whether apparent rank co-occurs with weak retention denominators. +""" + +from __future__ import annotations + +import argparse +import json +from pathlib import Path + +import numpy as np + +from observables_registry import ( + OBSERVABLES_CANONICAL, + OBSERVABLES_REGISTRY_VERSION, + compute_canonical, +) + + +OBS_NAMES = list(OBSERVABLES_CANONICAL.keys()) +PERT_NAMES = ["adjacent_swap", "block_shuffle", "large_gap_only", "uniform"] + + +def prime_gaps(n_gaps: int) -> np.ndarray: + """Return the first n_gaps prime gaps using a compact numpy sieve.""" + limit = max(100, int(n_gaps * (np.log(n_gaps + 10) + np.log(np.log(n_gaps + 10)) + 5))) + while True: + sieve = np.ones(limit + 1, dtype=bool) + sieve[:2] = False + for p in range(2, int(limit**0.5) + 1): + if sieve[p]: + sieve[p * p : limit + 1 : p] = False + primes = np.flatnonzero(sieve) + if len(primes) >= n_gaps + 1: + return np.diff(primes[: n_gaps + 1]).astype(float) + limit *= 2 + + +def gue_spacings(matrix_size: int, min_spacings: int, rng: np.random.Generator) -> np.ndarray: + """Generate unfolded GUE spacings by concatenating independent matrices.""" + parts = [] + edge = max(2, matrix_size // 10) + while sum(len(x) for x in parts) < min_spacings: + real = rng.standard_normal((matrix_size, matrix_size)) + imag = rng.standard_normal((matrix_size, matrix_size)) + h = real + 1j * imag + h = (h + h.conj().T) / (2.0 * np.sqrt(matrix_size)) + eigs = np.sort(np.linalg.eigvalsh(h).real) + bulk = eigs[edge:-edge] + gaps = np.diff(bulk) + mean = float(np.mean(gaps)) + if mean > 1e-15: + parts.append(gaps / mean) + return np.concatenate(parts)[:min_spacings].astype(float) + + +def perturb_adjacent_swap(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + idx = np.arange(0, len(out) - 1, 2) + chosen = idx[rng.random(len(idx)) < alpha] + tmp = out[chosen].copy() + out[chosen] = out[chosen + 1] + out[chosen + 1] = tmp + return out + + +def perturb_block_shuffle(gaps: np.ndarray, alpha: float, rng: np.random.Generator, block_size: int = 64) -> np.ndarray: + out = gaps.copy() + n_blocks = len(out) // block_size + if n_blocks <= 0: + return rng.permutation(out) if alpha > 0 else out + k = int(round(alpha * n_blocks)) + if k <= 0: + return out + for block in rng.choice(n_blocks, size=min(k, n_blocks), replace=False): + start = block * block_size + end = min(start + block_size, len(out)) + rng.shuffle(out[start:end]) + return out + + +def perturb_large_gap_only(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + idx = np.flatnonzero(out > np.median(out)) + k = int(round(alpha * len(idx))) + if k < 2: + return out + chosen = rng.choice(idx, size=min(k, len(idx)), replace=False) + vals = out[chosen].copy() + rng.shuffle(vals) + out[chosen] = vals + return out + + +def perturb_uniform(gaps: np.ndarray, alpha: float, rng: np.random.Generator) -> np.ndarray: + out = gaps.copy() + k = int(round(alpha * len(out))) + if k < 2: + return out + chosen = rng.choice(len(out), size=min(k, len(out)), replace=False) + vals = out[chosen].copy() + rng.shuffle(vals) + out[chosen] = vals + return out + + +PERTURB = { + "adjacent_swap": perturb_adjacent_swap, + "block_shuffle": perturb_block_shuffle, + "large_gap_only": perturb_large_gap_only, + "uniform": perturb_uniform, +} + + +def pca_summary(vectors: list[list[float]], labels: list[str]) -> dict: + matrix = np.array(vectors, dtype=float) + matrix = matrix - np.mean(matrix, axis=0, keepdims=True) + if min(matrix.shape) == 0 or np.max(np.abs(matrix)) <= 1e-15: + return { + "explained_variance": [], + "effective_rank": 0.0, + "centroid_cosine": {}, + "pc2": 0.0, + } + _, singular, _ = np.linalg.svd(matrix, full_matrices=False) + energy = singular * singular + if float(np.sum(energy)) <= 1e-15: + explained = np.zeros_like(energy) + effective_rank = 0.0 + else: + explained = energy / np.sum(energy) + pos = explained[explained > 1e-15] + effective_rank = float(np.exp(-np.sum(pos * np.log(pos)))) + + centroids = {} + for name in PERT_NAMES: + vals = np.array([v for v, label in zip(vectors, labels) if label == name], dtype=float) + if len(vals): + centroids[name] = np.mean(vals, axis=0) + + cosine = {} + for i, a_name in enumerate(PERT_NAMES): + for b_name in PERT_NAMES[i + 1 :]: + if a_name not in centroids or b_name not in centroids: + continue + a = centroids[a_name] + b = centroids[b_name] + denom = np.linalg.norm(a) * np.linalg.norm(b) + cosine[f"{a_name}_vs_{b_name}"] = float(np.dot(a, b) / denom) if denom > 1e-15 else 0.0 + + return { + "explained_variance": [float(x) for x in explained], + "effective_rank": effective_rank, + "centroid_cosine": cosine, + "pc2": float(explained[1]) if len(explained) > 1 else 0.0, + } + + +def analyze_sequence( + gaps: np.ndarray, + alphas: list[float], + n_trials: int, + n_baseline: int, + z_min: float, + rng: np.random.Generator, +) -> dict: + original = compute_canonical(gaps) + baseline_vals = {obs: [] for obs in OBS_NAMES} + for _ in range(n_baseline): + obs = compute_canonical(rng.permutation(gaps)) + for name in OBS_NAMES: + baseline_vals[name].append(obs[name]) + + baseline = {} + z = {} + denom = {} + for name in OBS_NAMES: + vals = np.array(baseline_vals[name], dtype=float) + mean = float(np.mean(vals)) + std = float(np.std(vals, ddof=1)) if len(vals) > 1 else 0.0 + baseline[name] = {"mean": mean, "std": std} + denom[name] = float(original[name] - mean) + z[name] = float(denom[name] / std) if std > 1e-15 else 0.0 + + stable_obs = [name for name in OBS_NAMES if abs(z[name]) >= z_min] + all_vectors = [] + screened_vectors = [] + labels = [] + profiles = [] + + for pert_name in PERT_NAMES: + for alpha in alphas: + trial_vals = {obs: [] for obs in OBS_NAMES} + for _ in range(n_trials): + perturbed = PERTURB[pert_name](gaps, alpha, rng) + obs = compute_canonical(perturbed) + for name in OBS_NAMES: + trial_vals[name].append(obs[name]) + means = {name: float(np.mean(trial_vals[name])) for name in OBS_NAMES} + retention = {} + for name in OBS_NAMES: + retention[name] = ( + float((means[name] - baseline[name]["mean"]) / denom[name]) + if abs(denom[name]) > 1e-12 + else 0.0 + ) + all_vector = [retention[name] for name in OBS_NAMES] + screened_vector = [retention[name] for name in stable_obs] + all_vectors.append(all_vector) + if len(stable_obs) >= 2: + screened_vectors.append(screened_vector) + labels.append(pert_name) + profiles.append( + { + "perturbation": pert_name, + "alpha": float(alpha), + "retention": retention, + "retention_vector": all_vector, + } + ) + + all_pca = pca_summary(all_vectors, labels) + screened_pca = pca_summary(screened_vectors, labels) if len(stable_obs) >= 2 else None + + return { + "n_gaps": int(len(gaps)), + "original": original, + "full_shuffle_baseline": baseline, + "denominator": denom, + "original_vs_shuffle_z": z, + "stable_observables": stable_obs, + "weak_observable_count": int(len(OBS_NAMES) - len(stable_obs)), + "profiles": profiles, + "pca_all_observables": all_pca, + "pca_stable_observables": screened_pca, + } + + +def summarize_replicates(items: list[dict]) -> dict: + def arr(path: tuple[str, ...]) -> np.ndarray: + vals = [] + for item in items: + x = item + for key in path: + if x is None: + break + x = x.get(key) + if isinstance(x, (int, float)): + vals.append(float(x)) + return np.array(vals, dtype=float) + + rank = arr(("pca_all_observables", "effective_rank")) + pc2 = arr(("pca_all_observables", "pc2")) + weak = np.array([item["weak_observable_count"] for item in items], dtype=float) + stable_rank = arr(("pca_stable_observables", "effective_rank")) + cos = arr(("pca_all_observables", "centroid_cosine", "adjacent_swap_vs_large_gap_only")) + + out = { + "n_replicates": len(items), + "rank_mean": float(np.mean(rank)) if len(rank) else 0.0, + "rank_std": float(np.std(rank, ddof=1)) if len(rank) > 1 else 0.0, + "pc2_mean": float(np.mean(pc2)) if len(pc2) else 0.0, + "pc2_std": float(np.std(pc2, ddof=1)) if len(pc2) > 1 else 0.0, + "weak_observable_count_mean": float(np.mean(weak)) if len(weak) else 0.0, + "stable_rank_mean": float(np.mean(stable_rank)) if len(stable_rank) else None, + "stable_rank_std": float(np.std(stable_rank, ddof=1)) if len(stable_rank) > 1 else 0.0, + "adjacent_vs_large_cosine_mean": float(np.mean(cos)) if len(cos) else 0.0, + "adjacent_vs_large_cosine_std": float(np.std(cos, ddof=1)) if len(cos) > 1 else 0.0, + } + if len(rank) > 1 and np.std(weak) > 1e-15 and np.std(rank) > 1e-15: + out["rank_vs_weak_count_corr"] = float(np.corrcoef(rank, weak)[0, 1]) + else: + out["rank_vs_weak_count_corr"] = 0.0 + return out + + +def build_prime_windows(max_n: int, n_reps: int) -> list[np.ndarray]: + total = max_n * n_reps + max_n + gaps = prime_gaps(total) + max_start = len(gaps) - max_n + starts = np.linspace(0, max_start, n_reps, dtype=int) + return [gaps[start : start + max_n].astype(float) for start in starts] + + +def run(args: argparse.Namespace) -> dict: + root_rng = np.random.default_rng(args.seed) + sizes = [int(x) for x in args.sizes.split(",") if x.strip()] + max_n = max(sizes) + alphas = [float(x) for x in np.linspace(args.alpha_min, args.alpha_max, args.n_alpha)] + + output = { + "experiment": "perturbation_rank_size_curve", + "question": "Does perturbation effective-rank stabilize with sample size under canonical observables?", + "observables_registry": OBSERVABLES_REGISTRY_VERSION, + "observables_used": OBS_NAMES, + "perturbations": PERT_NAMES, + "params": vars(args), + "alphas": alphas, + "domains": {}, + "summary": {}, + } + + prime_windows = build_prime_windows(max_n, args.n_replicates) + + print(f"observables_registry={OBSERVABLES_REGISTRY_VERSION}") + print(f"{'domain':<22} {'N':>6} {'rank':>7} {'PC2':>7} {'weak':>5} {'stable_rank':>11}") + + domain_builders = { + "primes_windows": lambda rep_rng, rep_i: prime_windows[rep_i], + "prime_shuffle_control": lambda rep_rng, rep_i: rep_rng.permutation(prime_windows[rep_i]), + "poisson": lambda rep_rng, rep_i: rep_rng.exponential(1.0, size=max_n), + "gue": lambda rep_rng, rep_i: gue_spacings(args.gue_matrix_size, max_n, rep_rng), + } + + for domain_name, builder in domain_builders.items(): + output["domains"][domain_name] = {} + output["summary"][domain_name] = {} + bases = [] + for rep_i in range(args.n_replicates): + rep_rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) + bases.append(builder(rep_rng, rep_i)) + + for n in sizes: + rows = [] + for rep_i, base in enumerate(bases): + rep_rng = np.random.default_rng(root_rng.integers(0, 2**63 - 1)) + res = analyze_sequence( + base[:n], + alphas=alphas, + n_trials=args.n_trials, + n_baseline=args.n_baseline, + z_min=args.z_min, + rng=rep_rng, + ) + res["replicate"] = rep_i + rows.append(res) + summary = summarize_replicates(rows) + output["domains"][domain_name][str(n)] = rows + output["summary"][domain_name][str(n)] = summary + stable_rank = summary["stable_rank_mean"] + stable_text = f"{stable_rank:>11.3f}" if stable_rank is not None else f"{'NA':>11}" + print( + f"{domain_name:<22} {n:>6} {summary['rank_mean']:>7.3f} " + f"{summary['pc2_mean']:>7.3f} {summary['weak_observable_count_mean']:>5.2f} " + f"{stable_text}" + ) + + out_path = Path(args.out) + out_path.parent.mkdir(parents=True, exist_ok=True) + with out_path.open("w") as f: + json.dump(output, f, indent=2) + print(f"saved {out_path}") + return output + + +def main() -> None: + parser = argparse.ArgumentParser() + parser.add_argument("--sizes", default="128,256,512,1024,2048") + parser.add_argument("--n-replicates", type=int, default=8) + parser.add_argument("--gue-matrix-size", type=int, default=180) + parser.add_argument("--n-alpha", type=int, default=5) + parser.add_argument("--alpha-min", type=float, default=0.1) + parser.add_argument("--alpha-max", type=float, default=0.9) + parser.add_argument("--n-trials", type=int, default=8) + parser.add_argument("--n-baseline", type=int, default=16) + parser.add_argument("--z-min", type=float, default=2.0) + parser.add_argument("--seed", type=int, default=20260506) + parser.add_argument("--out", default="tools/data/perturbation_rank_size_curve.json") + args = parser.parse_args() + run(args) + + +if __name__ == "__main__": + main() tokens used 140,057 Completato. Ho eseguito l’esperimento META/BOUNDARY sulla curva rank-vs-N con osservabili canonici e gate sui denominatori original-vs-shuffle. File prodotti: - Report: [agent_20260506_1941.md](/opt/MM_D-ND/tools/data/reports/agent_20260506_1941.md) - Script riusabile: [exp_perturbation_rank_size_curve.py](/opt/MM_D-ND/tools/exp_perturbation_rank_size_curve.py) - Dati: [perturbation_rank_size_curve.json](/opt/MM_D-ND/tools/data/perturbation_rank_size_curve.json) - Seme aggiornato: [seme.json](/opt/MM_D-ND/tools/data/seme.json) Risultante: il rank di perturbazione non è interpretabile senza gate sui denominatori. Poisson e shuffle-primi possono mostrare `rank_all ~1.8-2.0` con osservabili deboli; dopo gate `abs(z)>=2`, il rank torna vicino a 1. Nel perimetro testato, GUE non mostra un secondo asse stabile con osservabili canonici fino a N=2048. Verifiche eseguite: - `python -m py_compile tools/exp_perturbation_rank_size_curve.py` - validazione JSON di `seme.json` e `perturbation_rank_size_curve.json` - aggiornato `tools/data/reports/latest.md` al nuovo report.