Repository navigation
Accept only full-accuracy AL stages as converged - #99
Merged
Merged
Conversation
The KKT residual of `_abslog2_auglag` measures constraint violation, not how well a stage minimized its objective, so a stage solved at the loosest LSQR tolerance could be reported converged while its products were not covariant under a change of units (0.7% on Bai/bwm2000). `converged` now requires the accepting stage's last solve at the full-accuracy tolerance; an inexact stage that meets the test is followed by a stage at the floor tolerance. Within a stage, a full step that leaves the active pattern unchanged drops the tolerance to the floor, since the solve error is then the only remaining error; a fixed inexact tolerance dithers at its noise floor instead (40 solves on bwm2000, now 4). Iteration counts on sparse lognormal matrices at n = 31623 are within 5% of the previous ones. Adds a covariance test on the Brusselator Jacobian (Bai/bwm200) under power-of-two row scaling. Assisted-by: Claude Fable 5.1 <noreply@anthropic.com>
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #99 +/- ##
=======================================
Coverage 98.34% 98.34%
=======================================
Files 25 25
Lines 7538 7540 +2
=======================================
+ Hits 7413 7415 +2
Misses 125 125 ☔ View full report in Codecov by Harness. 🚀 New features to boost your workflow:
|
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
The KKT residual of
_abslog2_auglagmeasures constraint violation, not how well a stage minimized its objective, so a stage solved at the loosest LSQR tolerance could be reported converged while its products were not covariant under a change of units (0.7% on Bai/bwm2000).convergednow requires the accepting stage's last solve at the full-accuracy tolerance; an inexact stage that meets the test is followed by a stage at the floor tolerance.Within a stage, a full step that leaves the active pattern unchanged drops the tolerance to the floor, since the solve error is then the only remaining error; a fixed inexact tolerance dithers at its noise floor instead (40 solves on bwm2000, now 4).
Iteration counts on sparse lognormal matrices at n = 31623 are within 5% of the previous ones. Adds a covariance test on the Brusselator Jacobian (Bai/bwm200) under power-of-two row scaling.
Assisted-by: Claude Fable 5.1 noreply@anthropic.com