Repository navigation
Independent verification: quadratic screening brackets sharpen selected D/E diagnostics - #1669
Conversation
|
Batch review at Merged to Reviewed scope: Source-bound quadratic rational surrogate brackets for three selected moment roots, including surcharge contributions in every moment parameter. Written exponential-series comparison and independent arithmetic controls; narrows planning uncertainty without certifying finer endpoint signs. The pinned scoped command passed; the batch ledger contains its command, input hashes, environment and log. This incorporates the research/tooling contribution and does not change the selected κ or discharge the inherited multiplication-theorem assumptions. Batch review, attribution and evidence. Thank you for the contribution. |
2 participants
Result
Adds certified two-sided quadratic screening brackets for the three frozen
selected moment roots, retaining every modeled surcharge. The uncertainty
gap left by the linear
a < D/Ediagnostic shrinks by more than 419x fornative/conservative complex and more than 418x for bit.
This is a screening result, not a new kappa, endpoint-grid certification,
supplier improvement, speedup or selected-frontier change. The surrogate
gap is about 1.2e-8; containing the published 1e-24 endpoint intervals does
not independently certify their acceptance/rejection signs. It also cannot
resolve the much smaller native/conservative correction.
Exact scope and method
independent-moment.4fa6ddc953abae43610a0c49ace10c99.c191650c0fa8d2fa514e, refined toselected1094:three-ledgers:quadratic-majorant.2de86c77b4127d2f01f504e0c7c5e4db737caa3c.bit m=64,W=28416 with its surcharge. Profiles are read from pinned
ASSEMBLY/INDEPENDENT1094/bit EXPECTED files, not from headlines.
0fc329d0bec93ca7cd288eaea1295172e8789810;bit source head:
a5afc8c6169ea61ba7c51dfd3cade02cc7b13b46.For
M(a)=sum c_i exp(lambda_i*a), use exact bounds on its first and secondlogarithmic moments, E/F, and its maximum rate L. The factorial inequality
n!>=2*3^(n-2)gives1+t+t^2/2 <= exp(t) <= 1+t+t^2/[2(1-t/3)]on0<=t<3.The two rational gain surrogates bound the unique moment root, using the
correct outward bisection endpoints and the checked domain
Lhi*D/Elo<3.The surcharge is an exponential term and enters D, E, F and L.
No primary pricing/checking implementation is imported. The log primitive
uses an independently written positive atanh enclosure; that classical log
method is not claimed as novel. The different moment-root enclosure is
the rational surrogate comparison, not another full endpoint-series replay.
The first bounded computation succeeded; no pivot or large search was used.
Verification and files
Adds only
research/selected-quadratic-root-brackets/:README.md,TASK.json,SOURCES.json,VERIFY.json,RESULT.json,verify.py,test_verify.py. Frozen inputs and selected record are untouched.Observed on Windows/Python 3.12: metadata/pins PASS; registered replay PASS;
two unittest groups PASS with three explicit receipt mutations rejected.
Twelve mathematical/control groups include an independent exact fixture
M(a)=2^(a-1)with root 1, pole/invalid-domain/unbracketed-root rejection,and the surcharge's D/E/F/L effects. A separate read-only AI proof/code
review found no critical or important issue. Its minor maximum-rate coverage
point was addressed and replay rerun. A separate 100-digit Decimal root
calculation agreed with every bracket; that is a noncertifying cross-check,
not part of the exact acceptance argument.
Overlap, remaining obligations and credit
Entry survey and immediate refined concrete-term search preceded computation.
Final refresh at 2026-10-14T05:58Z covered 518 open PRs and 316 discussions
with complete pagination, plus local notes/ledger. Closest work is the linear
diagnostic #1293 at
85a361233431befe4b11cf878f7b1b77a244d473and TennisBowling's log-contractaudit #1405.
Existing complete selected endpoint audits are credited, not replaced.
No matching quadratic screening package was found; no novelty certificate
or code-diff coverage is claimed.
Claim: #1166 (comment).
The README/TASK give a bounded next question: source-bind one pending fully
charged profile and compare surrogate screening with its full replay before
claiming any timing benefit.
Profile realization, the surcharge's physical derivation, frame/owner/hosting,
precision/tape, all-size compiler, analytic transfer and bit H1-H5 obligations
remain inherited. No full selected cold replay, human/proof-assistant review,
supplier adoption or hosted-CI success is claimed. The unchanged fully modeled
curves are screened; no new whole-construction invoice is asserted.
Credit Rohan Arun (#1094),
Jeremy Kothe (#1087 and
mutable-leaver lineage), Hao Yan
(#858), DreamingOfClouds
(#589), eumemic and the
retained construction/review lineage. The prior diagnostic and TennisBowling
audit retain credit; their numerical routines are not imported or reused.
Prepared by Aaron Wang / navinor, with help from chatgpt, including
substantial research, implementation, writing and separate read-only AI
reviews. Apache-2.0.
Workflow feedback: selected endpoint replay is saturated, so the refined
question targeted a missing screening implication instead. No direct matching
open handoff was found. The requested custom worktree location was honored via
git worktree plus session.move because the native creator exposes no destination
argument. No dependencies were installed; the temporary worktree is to be
cleaned after publication at the donor's request.