10000 Instable test test_logistic_regressioncv_class_weights due to ConvergenceWarning · Issue #27568 · scikit-learn/scikit-learn · GitHub
[go: up one dir, main page]
Skip to content

Instable test test_logistic_regressioncv_class_weights due to ConvergenceWarning #27568

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Open
glemaitre opened this issue Oct 11, 2023 · 4 comments
Open

Instable test test_logistic_regressioncv_class_weights due to ConvergenceWarning #27568

glemaitre opened this issue Oct 11, 2023 · 4 comments

Comments

@glemaitre
Copy link
Member

When merging #26271, we added a lower tolerance in the following line:

https://github.com/scikit-learn/scikit-learn/pull/26721/files#diff-84dc33502e99175d160109f0ee16007297dc398d872a63c977b787287ec4ed00R752

However, it fails time to time on Ubuntu Atlas and the 32 bits docker CIs.
We also have a tolerance set to 1e-18 for the SAG and SAGA.

Should we increase the tolerance?

@lorentzenchr
@github-actions github-actions bot added the Needs Triage Issue requires triage label Oct 11, 2023
@davitacols

This comment was marked as spam.

Sign in to view
@lorentzenchr
Copy link
Member

@glemaitre You can try to use smaller solver tol for SAG/SAGA and larger rtol.

I often came across the same point: SAG/SAGA is very unstable and low precision. Its final convergence check relies on max_change <= tol * max_weight on the weights/coefficients alone whereas our high precision solvers all use something like infinity norm (max(abs(..=)) of the full (current) gradient.

@glemaitre
Copy link
Member Author

@davitacols Please stop posting GPT-like answer. It does not help here.

@glemaitre
Copy link
Member Author

@lorentzenchr Thanks I will have a look and try to reproduce it within a docker container with 32-bits arch.

@betatim betatim removed the Needs Triage Issue requires triage label Oct 19, 2023
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
4 participants
@betatim @glemaitre @lorentzenchr @davitacols
0