Skip to content

Prove sharp Gaussian radial budgets for every finite iterate identity - #7714

Open
aliromman wants to merge 2 commits into
CrocSwap:mainfrom
aliromman:codex/consecutive-result-047
Open

aliromman wants to merge 2 commits into
CrocSwap:mainfrom
aliromman:codex/consecutive-result-047

Conversation

@aliromman

@aliromman aliromman commented Nov 18, 2026 •

Copy link
Copy Markdown
Contributor

This scoped theorem gives the sharp radial budget for every prescribed finite Gaussian iterate identity. For a total Gaussian-pair map with exact sum, Euclidean error at most one from the positive half-transform and F^m=id, four must divide m. Let R be the first odd integer at least five that does not divide m. The exact minimum forward outward slack is sqrt((R^2+2R+3)/2)-sqrt((R^2+1)/2). No new kappa follows.

Read the proof, receipt, local map and checker and credits.

At m=20 the optimum is sqrt(33)-5; at m=140 it is sqrt(51)-sqrt(41). The known m=4 value sqrt(19)-sqrt(13) remains prior work. Finite iterate budgets approach 1/sqrt(2) as their compatible odd ring periods increase, while every fixed finite m has a strict gap above that limit. The requirement is an iterate identity, not that m be the least positive global period. The constructed map's least period can be smaller.

The lower proof isolates the first incompatible ring between two invariant finite squares. It assumes no symmetry. The upper construction combines directed single-ring cycles with shifted annular pairs and handles a separate central case when three does not divide m. It is total on every sum fiber and both half-grid colors, has an explicit inverse, and attains the same sharp bound for forward, inverse outward and absolute radial change. Its local point rule uses no unbounded matching table. A four-cycle supplies exact attainment after the all-input upper proof is established.

This resolves the general finite-iterate optimum left open in PR #7699. No native tariff, selected caller, tensor commutation, inverse-by-conjugation contract, scratch restoration or complete recursive invoice is supplied. The selected kappa stays 0.00363005817, with all five assumptions retained. Exact least period, extra state, larger blocks and changed sum/error contracts remain outside scope.

With CPython 3.12 and its standard library:

python3 -B scripts/research_package.py --package research/gaussian-finite-iterate-optimum

Both registered normal and optimized commands print:

PASS_GAUSSIAN_FINITE_ITERATE_OPTIMUM
1024 iterate budgets; 23 complete grids; 41032 half-grid points
2424 lower-boundary edges; 19208 full Gaussian inputs; all controls pass

Twenty-three complete grids cover 20 critical cutoffs from five through 61, including both central choices at R=5,7,9. A separate arc generator agrees with the local map. Direct traversal checks every cycle and its order; all inverses, unit edges, both colors, exact radial inequalities and attaining cycles pass. The 2424 exterior edges of the critical square pairs meet the lower threshold. Another 19208 Gaussian inputs cover all component parities and eight iterate choices. All arithmetic is integral with certified square-root comparisons.

Omitting the central annular case fails m=4,20,140. Leaving the first incompatible ring as a single ring fails the required iterate. Reusing the unchanged fourth-power construction exceeds the new m=20 budget. The zero-sum control has exact period four. The all-m and all-input claims rest on the written proof; finite cycles and grids are corroboration.

Normal/optimized replay, committed bytes and modes, source history, export and sealed Git-free replay pass. Hosted CI did not certify this package. The plan job failed in the shared test_current_ci_section_matches_selected_record: the base guide names frontier6030 while the selected record requires make frontier7000-verify. The log reports 760 tests, one failure and 24 skipped; package replay was skipped. This package changes no shared CI file. Local replay remains the scientific evidence. Full predecessor suites, selected cold replay, native/physical replay, Lean, Docker, independent human and formal-kernel review were not run. The adapted checker is not an independent source review.

Credit Jerry Chen (@CubeJerry) for the half-grid, annular matching and sharp square-to-swap construction in PR #7570, head 053c42c68165cf24fd4e8bc33c2b20d6ac3d399b: all-input proof. The annular formula and connector cycle are adapted to new starting radii; its m=4 optimum is retained.

Ali Romman's PR #7699, head 2a5dcb314ccdfeed24bd86ec1f56f519eead56c1, supplies the ring classification and finite-order lower framework and adapted exact comparison routines. Its source credits retain the global-optimum, provider, earlier Gaussian and Nick Petalas exact-sum contributions. Prepared by Ali Romman with substantial OpenAI Codex assistance, Apache-2.0.

Base: current main e54523517d0f03e58352e647c403a5af3e525a57, priority revision 8. The claim records the changed finite-iterate question. Live handoffs, current source heads, source-reference intake, phrase overlap and the closed delta were checked. Sparse local and unpublished work remain coverage limits. Open next questions require an actual caller contract before further physical or exact-period work. No follow-up is reserved.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant