diff --git a/research/four-wire-parity-suffix/README.md b/research/four-wire-parity-suffix/README.md new file mode 100644 index 0000000000..a34451d071 --- /dev/null +++ b/research/four-wire-parity-suffix/README.md @@ -0,0 +1,194 @@ +# The sharp four-wire forward-suffix envelope + +This is a scoped algebraic proof with exact finite corroboration. For the +retained triangular Toeplitz parity compiler on any complex symbol +`f(z)=1+bz+az²+cz³`, the maximum forward-suffix infinity norm is exactly + +```text +M(a,b,c) = 1 + max(|a|+|b|+|c|, + |b|+|a-b²|+max(|c|,|c-ab|)). +``` + +The maximum includes both endpoints and every complete-shear cut of every +legal order of the two disjoint children. Consequently, for every real +`eta >= 0`, + +```text +sup_{|a|+|b|+|c| <= eta} M(a,b,c) = 1+eta+eta². +``` + +Equality holds exactly when `a=c=0` and `|b|=eta`. The statement has no +small-tail or sign restriction. It solves the explicit four-wire optimization +suggested in [#4831](https://github.com/CrocSwap/integer-mult-bounds/pull/4831), +whose all-depth theorem concerns only binomial symbols. + +For `00`, equality forces `t=eta` and +`B(A+B)=t²`. Since both `B` and `A+B` are at most `t`, this forces +`B=t`, hence `A=C=0`. Conversely, for every complex `b` of modulus `eta`, +the formula gives `M(0,b,0)=1+eta+eta²`. If `eta=0`, all coefficients vanish +and the same classification holds. This proves the sharp envelope and all +its extremizers. The comparison with #4731 follows by subtracting the +geometric-series envelope. No predecessor scientific executable is replayed. + +## Exact finite corroboration + +The standard-library verifier constructs the literal root and child macros +by truncated Gaussian-rational polynomial division. It executes both child +orders, multiplies the prefix and suffix matrices at every cut, and compares +the maximum with the separate closed formula. It also checks the bound and +the extremizer criterion for every fixture. Matrix entries are exact pairs +of rational real and imaginary parts. + +Norms are exact sums of square roots of rationals, reduced to rational +multiples of squarefree radicals. Equality compares these canonical sums; +ordering uses outward rational intervals built with integer square roots. +Unresolved ordering raises an error after 256 fractional bits. No binary +floating-point comparison or tolerance decides acceptance. + +The complete fixture grid uses zero and the five phases +`1,-1,i,-i,(3+4i)/5` at amplitudes `1/4,1/2`, independently for `a,b,c`. +Thus it includes non-real inputs whose derived coefficients have irrational +moduli. All 1,331 grid fixtures and 13,310 order-labelled cuts pass; exactly +11 fixtures attain their own tail envelope, and 605 give a larger maximum +under even-first execution than under odd-first execution. These cut counts +retain duplicated matrices at different chronological cuts. Ten controls +pass, including the two invalid-coefficient cases. Controls reject omitted +Schur and root-left macros, wrong chronology, endpoint-only bounds, applying +the binomial value to all symbols, omitting the necessary child order, and +invalid coefficients. Separate arithmetic controls exercise irrational +comparison and exact Pythagorean modulus equality. + +From the repository root, with CPython 3.12: + +```sh +/Users/nick/git/integer-mult-bounds/.venv/bin/python3.12 -I -W error -X dev -B research/four-wire-parity-suffix/verify.py +/Users/nick/git/integer-mult-bounds/.venv/bin/python3.12 -I -O -W error -X dev -B research/four-wire-parity-suffix/verify.py +/Users/nick/git/integer-mult-bounds/.venv/bin/python3.12 -B scripts/research_package.py --package research/four-wire-parity-suffix +/Users/nick/git/integer-mult-bounds/.venv/bin/python3.12 -B scripts/check_research_commit.py --package research/four-wire-parity-suffix --check-source-history +``` + +Another CPython 3.12 executable may replace the local interpreter path. +The expected marker is `PASS_FOUR_WIRE_PARITY_SUFFIX`. The script writes +nothing and emits a generated receipt binding the five package-file hashes. +The registered replay runs normal and optimized warning-strict checks. +Universal complex claims rely on the written proof, rather than finite +extrapolation. Independent human and formal proof-kernel review are unclaimed. + +## Limits and code constraints + +The result concerns exactly four wires and complete-shear boundaries. It +does not bound cuts inside a dirty-shear implementation, arbitrary larger +symbols, rounded arithmetic, complete Gaussian stages, tape costs or the +selected supplier. All five framework assumptions remain at their existing +strength. No selected cold replay, Lean or Docker was run. + +The verifier has no recursion, dynamic imports, subprocesses or network +access. Every function is under 60 lines. The fixed grid has 1,331 fixtures, +two four-macro words per fixture and five cuts per word. Matrices are always +four by four. Radical factorization accepts integers at most `2^40`, uses at +most `2^20` trial divisors and at most 41 divisions per divisor. Exact norm +ordering makes at most five interval refinements. Python's managed allocation +and arbitrary-precision temporaries are a documented Power of Ten deviation; +this is a bounded finite checker, with no public unbounded input interface. +Internal helpers consume only the fixed grid and fixed four-wire structures. +Warning-strict and optimized runs are the available interpreter diagnostics. + +## Open next questions + +1. Determine the sharp forward-suffix envelope for general eight-wire symbols. + This would give the next precision-interface bound beyond this theorem. + Start with the child-cut row formulas and preserve correlations between + the even child and its inverse. A finite candidate grid would not prove + the complex optimum, and binomial optimality at larger sizes is unclaimed. +2. Bind one actual four-wire shear implementation to this macro envelope. + Start with its complete literal gate sequence and rounding convention. + Internal gates, coefficient delivery and the complete physical bill + remain additional obligations. + +## Sources and credit + +This contribution was produced by Nick Petalas. Apache-2.0. + +The source question is #4831's explicit four-wire optimization. That package's +binomial values, all-depth recurrence and attaining family remain credited +predecessor results. Brian / JoyboyBrian #4731 supplies the common macro-cut +contract and four-operator envelope. Vlado Vojdanovski #1673 supplies the +parity compiler, retained by Cinna Davis #4100. James Reed #4699 supplies +one-level tail accounting; Ryanson Jonathan and unpaidmoon #4371 supply +the earlier conditioning question. The new contribution is the full complex +four-wire formula, its sharp uniform upper bound and its extremizers. + +Canonical base is `e54523517d0f03e58352e647c403a5af3e525a57`, priority +revision 8, selected #7000 unchanged. The catalogue's older selected context +was recognized; current selected source bindings passed. [TASK.json](TASK.json) +records the changed scope from audit to an original precision-interface lemma. +[SOURCES.json](SOURCES.json) pins historical provenance only. Direct replay +requires this package alone; no predecessor numerical claim is rechecked. +The [advance claim](https://github.com/CrocSwap/integer-mult-bounds/discussions/1166#discussioncomment-19595240) +states this same scope. Live PR, main-branch README, merged-work and local +reservation checks found no resolving general four-wire suffix formula. +This textual overlap search does not cover unpublished work or every diff. diff --git a/research/four-wire-parity-suffix/SOURCES.json b/research/four-wire-parity-suffix/SOURCES.json new file mode 100644 index 0000000000..2a99e1345a --- /dev/null +++ b/research/four-wire-parity-suffix/SOURCES.json @@ -0,0 +1,36 @@ +{ + "schema_version": 1, + "base_commit": "e54523517d0f03e58352e647c403a5af3e525a57", + "files": [ + { + "path": "certificates/selected-result.json", + "usage": "provenance", + "git_blob": "75d77b328ea02ded535e7f23aee5b1e55d0bb1a8", + "sha256": "154c3d30455d778fd3343b6b4851f0fa4665f0f7625b0adb3675e2bafbbf6bf1" + }, + { + "path": "docs/research/agent-task-catalog.json", + "usage": "provenance", + "git_blob": "3404fbdd5e7529a0674eef26114b5071e4586cec", + "sha256": "ead6cdc9d0af6d58e8a49744d8ec8c7c15e13d2ba93c276c434bcf9b51b93a81" + }, + { + "path": "research/triangular-parity-binomial-suffix/README.md", + "usage": "provenance", + "git_blob": "40be9dc4c7cf7a8e0b1e4f29267adeecdb4bddd8", + "sha256": "9c92249119c31c97b970975ef28ea8a23263acd8e5d5b95f083b71af93b2d5f1" + }, + { + "path": "research/triangular-parity-uniform-prefix/README.md", + "usage": "provenance", + "git_blob": "6e64fb48fbd33b9742f0ba81de77bfb661b80403", + "sha256": "dd0259414df08b5b462cec450df5a0618c30ba112ddf87b2acbb2e39a1c69414" + }, + { + "path": "research/unconditional-batched-crt/vendor/SHALLOW-TOEPLITZ.md", + "usage": "provenance", + "git_blob": "bc4349b2daea548ffa09e1f38a31f1df86b0b7be", + "sha256": "b384b93f3fb4a8c97da2e617ebc14d55cebb0a739070e7f8dbb66f63e24e9310" + } + ] +} diff --git a/research/four-wire-parity-suffix/TASK.json b/research/four-wire-parity-suffix/TASK.json new file mode 100644 index 0000000000..39505da0d4 --- /dev/null +++ b/research/four-wire-parity-suffix/TASK.json @@ -0,0 +1,89 @@ +{ + "schema_version": 1, + "task_id": "uniform-transfer-audit", + "priority_revision": 8, + "base_commit": "e54523517d0f03e58352e647c403a5af3e525a57", + "original_card": { + "id": "uniform-transfer-audit", + "question": "Can one of the selected record's five inherited assumptions be proved at its actual caller?", + "starting_points": [ + "docs/research/current-priorities.md", + "docs/research/research-family-status.md", + "certificates/selected-result.json", + "docs/research/pr-intake-evidence/pr5599-completion/PROOF-REVIEW.md" + ], + "methods": [ + "Construct and independently check one source-bound instance of the stated first experiment.", + "Seek an exact counterexample or target-specific complete-budget obstruction to that same proposed change." + ], + "acceptance_gate": "Identify the proposition and what would change downstream; do not equate a finite replay with a universal proof.", + "first_step": { + "command": [ + "python3", + "-B", + "scripts/research_selected_baseline.py" + ], + "expected": "PASS_RESEARCH_BASELINE_BINDINGS_ONLY", + "inspect": "Read revision 8, the selected review and live exact-source handoffs. Then follow this card's bounded first_experiment.", + "scope": "Source-bound orientation only, not a supplier replay or theorem proof." + }, + "first_experiment": "Choose one assumption ID (native_all_columns, owned_erasure, compiler, precision_tape, analytic_recovery), one quantified implication and exact dependency closure. Check current handoffs, including #5625 row reentry and #5707 root-prefix lowering, without treating their partial scope as the whole theorem. Prove or falsify that implication with a caller-bound argument; stop at the chosen obligation. Finite tests or root digits do not discharge an all-size statement.", + "bottleneck_link": "Judge the actual current supplier contract and complete signed target/paid bill; frozen-profile ceilings require an applicability proof.", + "allowed_variations": [ + "One exact source head, helper/cut region or quantified proposition and changed hypothesis; check live handoffs first. Different seeds are not different scopes." + ], + "scope_change_triggers": [ + "Changing source head, helper/cut domain, operation alphabet or caller boundary without recording the new scope.", + "Treating an archived proposal or local numerical pass as complete supplier admission." + ], + "overlap_terms": [ + "native_all_columns", + "owned_erasure", + "compiler", + "precision_tape", + "analytic_recovery" + ], + "existing_checks": [ + { + "scope": "PR #4100: proposed unconditional construction and finite checks, without independent acceptance.", + "evidence": "research/unconditional-batched-crt/README.md", + "remaining": "Independent proof audit of one caller-bound all-size implication; all other premises remain explicit.", + "status": "author-checked", + "source_commit": "337d04939571a637aeab54a0f58f54fde811663a", + "selected_record_sha256": "bb2eaa58647b353f88d0aec2eafb9d30df1be53589231b23e12ceef3e4bad129" + }, + { + "scope": "Selected PR5599: complete helper-subspace construction and independent caller replay; five framework assumptions remain.", + "evidence": "docs/research/pr-intake-evidence/pr5599-completion/PROOF-REVIEW.md", + "remaining": "The next experiment must identify uncovered source-specific work. Successor archive intake does not transfer this admission.", + "status": "reviewed", + "source_commit": "09dd0965711ca7eeb169057a420abab1c9052137", + "selected_record_sha256": "2a0220e866417774606e0f991e41b9c1d26b1f43b97857238cf478dbc7d330b5" + } + ] + }, + "scope_status": "changed", + "actual_scope": "Prove the exact forward-suffix infinity norm for every complex four-wire triangular parity symbol and the sharp 1+eta+eta^2 envelope with all extremizers.", + "scope_change_reason": "Take the explicit four-wire optimization in PR4831 Open next questions. This is an original scoped precision-interface result, not an audit, predecessor replay or discharge of a selected framework assumption.", + "bottleneck_link": "Sharper macro-suffix norms constrain error amplification under the retained precision_tape interface; internal gates, an actual caller and its complete paid bill remain separate.", + "budget_minutes": 60, + "stop_rule": "Ship this first exact four-wire formula and sharp envelope after bounded exact corroboration. Leave general eight-wire and gate-level questions open.", + "outcome": "completed", + "evidence": "README contains the universal complex proof; verify.py enumerates every literal macro cut for its complete fixed Gaussian-rational grid and runs semantic controls.", + "author": "Nick Petalas", + "ai_assistance": "Independent human and formal proof-kernel review are unclaimed.", + "open_questions": [ + { + "question": "What is the sharp general eight-wire forward-suffix envelope?", + "why_it_matters": "It extends the one-sided precision budget beyond this exact four-wire theorem.", + "first_step": "Derive correlated child-cut row formulas and test them against exact candidate symbols.", + "blocked_by": "Independent norm products lose correlations and a finite grid cannot prove the complex optimum." + }, + { + "question": "Can an actual four-wire shear implementation use this macro envelope?", + "why_it_matters": "A macro bound alone does not price internal rounding or physical work.", + "first_step": "Bind one literal shear sequence and rounding convention.", + "blocked_by": "Internal gate bounds, coefficient delivery and the complete physical bill remain missing." + } + ] +} diff --git a/research/four-wire-parity-suffix/VERIFY.json b/research/four-wire-parity-suffix/VERIFY.json new file mode 100644 index 0000000000..e5a5490ea3 --- /dev/null +++ b/research/four-wire-parity-suffix/VERIFY.json @@ -0,0 +1,38 @@ +{ + "schema_version": 1, + "status": "ready", + "python": "3.12", + "requirements": [], + "inputs": [], + "commands": [ + { + "argv": [ + "{python}", + "-I", + "-W", + "error", + "-X", + "dev", + "-B", + "{package}/verify.py" + ], + "timeout_seconds": 120, + "checks": "Exact Gaussian-rational macro execution, all-cut endpoint identities, radical norm formula, sharp envelope, extremizers and semantic controls." + }, + { + "argv": [ + "{python}", + "-I", + "-O", + "-W", + "error", + "-X", + "dev", + "-B", + "{package}/verify.py" + ], + "timeout_seconds": 120, + "checks": "Exact Gaussian-rational macro execution, all-cut endpoint identities, radical norm formula, sharp envelope, extremizers and semantic controls." + } + ] +} diff --git a/research/four-wire-parity-suffix/verify.py b/research/four-wire-parity-suffix/verify.py new file mode 100644 index 0000000000..fc2577271a --- /dev/null +++ b/research/four-wire-parity-suffix/verify.py @@ -0,0 +1,256 @@ +#!/usr/bin/env python3 +"""Exact four-wire parity suffix corroboration. + +Produced by Nick Petalas. Apache-2.0. +""" + +from fractions import Fraction as F +from hashlib import sha256 +from itertools import product +import json +from math import isqrt +from pathlib import Path + +ZERO = (F(0), F(0)) +ONE = (F(1), F(0)) + + +def require(condition, message): + if not condition: + raise AssertionError(message) + + +def add(x, y): + return (x[0] + y[0], x[1] + y[1]) + + +def sub(x, y): + return (x[0] - y[0], x[1] - y[1]) + + +def mul(x, y): + return (x[0] * y[0] - x[1] * y[1], + x[0] * y[1] + x[1] * y[0]) + + +def radical(integer): + """Return c,d with sqrt(integer)=c*sqrt(d), d squarefree.""" + if not 0 <= integer <= 2 ** 40: + raise ValueError("radicand capacity") + coefficient, residual, remainder = 1, 1, integer + if not remainder: + return 0, 1 + for divisor in range(2, isqrt(integer) + 1): + if divisor * divisor > remainder: + break + exponent = 0 + for _ in range(41): + if remainder % divisor: + break + remainder //= divisor + exponent += 1 + coefficient *= divisor ** (exponent // 2) + if exponent % 2: + residual *= divisor + if remainder > 1: + residual *= remainder + require(coefficient * coefficient * residual == integer, + "exact radical decomposition") + return coefficient, residual + + +def magnitude(z): + square = z[0] * z[0] + z[1] * z[1] + coefficient, residual = radical(square.numerator * square.denominator) + return () if not coefficient else ((residual, F(coefficient, square.denominator)),) + + +def total(*expressions): + terms = {} + for expression in expressions: + for radicand, coefficient in expression: + terms[radicand] = terms.get(radicand, F(0)) + coefficient + return tuple(sorted((r, c) for r, c in terms.items() if c)) + + +def compare(x, y): + difference = total(x, tuple((r, -c) for r, c in y)) + if not difference: + return 0 + for bits in (16, 32, 64, 128, 256): + lower = upper = F(0) + scale = 2 ** bits + for radicand, coefficient in difference: + square = radicand * scale * scale + floor = isqrt(square) + ceiling = floor + (floor * floor != square) + lo, hi = F(floor, scale), F(ceiling, scale) + lower += coefficient * (lo if coefficient > 0 else hi) + upper += coefficient * (hi if coefficient > 0 else lo) + if lower > 0: + return 1 + if upper < 0: + return -1 + raise ArithmeticError("exact comparison did not separate within capacity") + + +def largest(expressions): + require(1 <= len(expressions) <= 16, "finite maximum capacity") + answer = expressions[0] + for expression in expressions[1:]: + if compare(expression, answer) > 0: + answer = expression + return answer + + +def identity(): + return tuple(tuple(ONE if i == j else ZERO for j in range(4)) + for i in range(4)) + + +def multiply(a, b): + rows = [] + for i in range(4): + row = [] + for j in range(4): + value = ZERO + for k in range(4): + value = add(value, mul(a[i][k], b[k][j])) + row.append(value) + rows.append(tuple(row)) + return tuple(rows) + + +def norm(matrix): + return largest(tuple(total(*(magnitude(z) for z in row)) for row in matrix)) + + +def elementary(updates): + rows = [list(row) for row in identity()] + for target, source, coefficient in updates: + rows[target][source] = add(rows[target][source], coefficient) + return tuple(map(tuple, rows)) + + +def compiler(a, b, c, even_first): + """Derive the literal parity macros by truncated polynomial division.""" + for z in (a, b, c): + if len(z) != 2 or any(not isinstance(x, F) or abs(x) > 2 for x in z): + raise ValueError("bounded Gaussian-rational coefficient required") + even, odd = (ONE, a), (b, c) + quotient = [ZERO, ZERO] + for i in range(2): + quotient[i] = odd[i] + for j in range(1, i + 1): + quotient[i] = sub(quotient[i], mul(even[j], quotient[i - j])) + schur = sub(even[1], mul(odd[0], quotient[0])) + right = elementary(((2, 1, quotient[0]),)) + left = elementary(((1, 0, quotient[0]), (3, 0, quotient[1]), + (3, 2, quotient[0]))) + children = (elementary(((2, 0, even[1]),)), + elementary(((3, 1, schur),))) + return (right, *children[::1 if even_first else -1], left) + + +def cut_products(word): + prefixes = [identity()] + for gate in word: + prefixes.append(multiply(gate, prefixes[-1])) + suffixes = [identity()] * 5 + for i in range(3, -1, -1): + suffixes[i] = multiply(suffixes[i + 1], word[i]) + return tuple(zip(prefixes, suffixes)) + + +def predicted(a, b, c): + endpoint = total(magnitude(a), magnitude(b), magnitude(c)) + residual = sub(c, mul(a, b)) + schur = sub(a, mul(b, b)) + transported = total(magnitude(b), magnitude(schur), + largest((magnitude(c), magnitude(residual)))) + return total(magnitude(ONE), largest((endpoint, transported))) + + +def target(a, b, c): + coefficients = (ONE, b, a, c) + return tuple(tuple(coefficients[i - j] if i >= j else ZERO + for j in range(4)) for i in range(4)) + + +def fixture(a, b, c): + expected = target(a, b, c) + maxima = [] + for order in (True, False): + cuts = cut_products(compiler(a, b, c, order)) + for prefix, suffix in cuts: + require(multiply(suffix, prefix) == expected, "all-input endpoint at cut") + maxima.append(largest(tuple(norm(q) for _, q in cuts))) + observed = largest(tuple(maxima)) + require(compare(observed, predicted(a, b, c)) == 0, + "exact full-complex per-symbol formula") + tail = total(magnitude(a), magnitude(b), magnitude(c)) + require(all(r == 1 for r, _ in tail), "rational fixture tail") + t = sum((v for _, v in tail), F(0)) + envelope = ((1, 1 + t + t * t),) + equality = compare(observed, envelope) == 0 + require(compare(observed, envelope) <= 0, "tail-budget envelope") + require(equality == (a == ZERO and c == ZERO), "complete extremizer criterion") + return 10, equality, compare(maxima[0], maxima[1]) > 0 + + +def controls(): + half, quarter = (F(1, 2), F(0)), (F(1, 4), F(0)) + word = compiler(ZERO, half, ZERO, True) + expected = target(ZERO, half, ZERO) + require(cut_products((word[0], word[1], identity(), word[3]))[-1][0] + != expected, "missing Schur child detected") + require(cut_products((*word[:3], identity()))[-1][0] + != expected, "missing root left macro detected") + require(cut_products((*word[1:], word[0]))[-1][0] + != expected, "wrong chronology detected") + require(compare(predicted(ZERO, half, ZERO), norm(expected)) > 0, + "endpoint-only bound rejected") + require(compare(predicted(half, ZERO, ZERO), ((1, F(7, 4)),)) < 0, + "binomial value is not every symbol's exact value") + require(fixture(sub(ZERO, quarter), half, ZERO)[2], + "odd-first-only schedule misses the maximum") + require(compare(magnitude((F(1), F(1))), ((1, F(7, 5)),)) > 0, + "nonrational square root comparison") + require(compare(magnitude((F(3, 5), F(4, 5))), magnitude(ONE)) == 0, + "Pythagorean modulus equality") + rejected = 0 + for bad in ((3, F(0)), (F(3), F(0))): + try: + compiler(bad, ZERO, ZERO, True) + except ValueError: + rejected += 1 + require(rejected == 2, "coefficient type and capacity failures") + return 10 + + +def main(): + phases = ((F(1), F(0)), (F(-1), F(0)), (F(0), F(1)), + (F(0), F(-1)), (F(3, 5), F(4, 5))) + values = (ZERO,) + tuple(mul((scale, F(0)), phase) + for scale in (F(1, 4), F(1, 2)) for phase in phases) + cuts = equalities = order_gaps = fixtures = 0 + for a, b, c in product(values, repeat=3): + count, equality, gap = fixture(a, b, c) + fixtures += 1 + cuts += count + equalities += equality + order_gaps += gap + controls_count = controls() + require(fixtures == 11 ** 3 and cuts == 10 * fixtures, "complete finite grid") + require(equalities == len(values), "every and only binomial extremizer") + files = sorted(Path(__file__).resolve().parent.iterdir()) + hashes = {p.name: sha256(p.read_bytes()).hexdigest() for p in files if p.is_file()} + print(json.dumps({"status": "PASS_FOUR_WIRE_PARITY_SUFFIX", "fixtures": fixtures, + "cuts": cuts, "extremizers": equalities, + "order_sensitive_fixtures": order_gaps, + "controls": controls_count, "errors": 0, + "sha256": hashes}, sort_keys=True)) + + +if __name__ == "__main__": + main()