Repository navigation
[MRG] Stabilize simplex projection against large common offsets - #880
Open
AHMETHAKANBEZIR1 wants to merge 9 commits into
Open
AHMETHAKANBEZIR1 wants to merge 9 commits into
AHMETHAKANBEZIR1 wants to merge 9 commits into
Conversation
Co-authored-by: Codex <noreply@openai.com>
Co-authored-by: Codex <noreply@openai.com>
…tion-stability # Conflicts: # RELEASES.md
…tion-stability # Conflicts: # RELEASES.md
…tion-stability # Conflicts: # RELEASES.md
This was referenced Nov 10, 2026
…tion-stability # Conflicts: # RELEASES.md
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Fixes #879. Simplex projection is invariant to a common offset per column. Center floating inputs before sorting and thresholding so large equal float32/float64 inputs retain unit mass rather than producing all zeros. Integer inputs retain their existing path. This does not change sparse-projection handling or validation for nonpositive simplex radii.
Validation
Current head
1cc4127includes masterf48a63a. OnlyRELEASES.mdconflicted; both release entries are retained. The PR production and regression files are byte-for-byte unchanged from its previous head.Complete utility module and doctests: NumPy 127 passed, 1 skipped; NumPy/PyTorch 2.9.1 CPU 193 passed, 1 skipped; NumPy/TensorFlow 161 passed, 33 skipped. Each process asserts the actual backend list and source checkout. Torch and TensorFlow run in separate environments. Skips are reported, not counted as passes.
The updated master's native extensions were rebuilt locally with the existing MSVC/Cython toolchain; the results above were repeated against that build. No native solver correctness or performance claim follows from the utility tests.
All applicable configured pre-commit hooks for the PR files and diff checks pass.
A private integration of [MRG] Fix sparse simplex projection with oversized budgets #876, [MRG] Handle both endpoints in sample-ratio splitting #878 and [MRG] Stabilize simplex projection against large common offsets #880 on the same master passes the utility module/doctests: NumPy 145 passed, 1 skipped; NumPy/PyTorch 229 passed, 1 skipped; NumPy/TensorFlow 197 passed, 33 skipped. That local integration branch is not published and is not part of this PR.
Windows/Python 3.12.14, CPU only. The full package suite, current full documentation/gallery, GPU, JAX and CuPy were not run locally. Current-head CircleCI documentation build 4314 completed successfully, with its commit verified against this PR head; all three reported checks now pass. This is an upstream full documentation build, not a local gallery run.
Contribution disclosure
Prepared, investigated and tested autonomously by OpenAI Codex on behalf of AHMETHAKANBEZIR1. No independent human review has been performed.
15 November 2026 update (dbe2d6c). Merged master 3367480 and resolved the RELEASES.md conflict by keeping both entries. The PR's implementation and regression tests are unchanged. The utility and circle-solver test modules passed on Windows CPU (144 NumPy tests and 221 NumPy/PyTorch tests). The new CircleCI documentation checks are pending at the time of this update.