Repository navigation
lasso: give fresh S&P downloads a decade of headroom - #37
Merged
Merged
Conversation
A clean-clone reproduction (fresh uv cache, uv run --locked, both data sets downloaded) compared every S&P corner rather than every 10th, and on that download the QP baseline missed by up to 8.8e-10: inside the 1e-9 tolerance, but with little margin. The path's own KKT residual was 3.0e-12. Allow 1e-8 on data that are not the snapshot (1e-10 on it, as before), and say in the note that the QP drifts by up to 9e-10 and what the tests allow. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
tschm
added a commit
that referenced
this pull request
Oct 14, 2026
CI: #37 and #38 were merged within a minute, and #37's older build finished publishing 23 seconds after #38's, so the paper and arxiv branches went back to the older commit. The publish job now runs in one concurrency group without cancelling, and a build that is no longer of main's tip skips publishing, so the newest merge's files always win. lasso.tex: the note uses British spelling throughout (optimisation, factorisation, centred), so 'MIT licence', the noun. A spelling pass found nothing else. Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Follow-up to #36. #36 was merged before this commit reached it.
A clean-clone reproduction, done the way the journal will do it (fresh clone, empty uv cache,
uv run --locked, both data sets downloaded), compared every S&P corner. The test compares only every 10th. On that download the QP baseline's worst corner gap was 8.8e-10: inside #36's 1e-9 tolerance, but with little margin. The path's own KKT residual was 3.0e-12.From the same clean-clone run:
🤖 Generated with Claude Code