Align ridge alpha path computation with glmnet #972
+51
−30
Conversation
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
- Compute data-dependent alpha_max for pure ridge (l1_ratio=0) using the null model gradient divided by a small surrogate (0.001), instead of hardcoding alpha_max=10. Falls back to 10 when the gradient is zero (e.g. constant X after centering). - Default min_alpha_ratio is now 1e-2 when n_samples < n_features, matching glmnet's convention. Remains 1e-6 when n_samples >= n_features. - Extend test_alpha_path to cover the pure ridge case (l1_ratio=0). Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
MatthiasSchmidtblaicherQC
requested review from
MarcAntoineSchmidtQC,
jtilly and
stanmart
as code owners
MatthiasSchmidtblaicherQC
requested review from
Copilot
and removed request for
MarcAntoineSchmidtQC,
jtilly and
stanmart
Copilot started reviewing on behalf of
MatthiasSchmidtblaicherQC
February 8, 2026 16:32
View session
There was a problem hiding this comment.
Pull request overview
This PR updates how glum builds the regularization (alpha) path for ridge regression to better match glmnet, and adjusts the default path length heuristic based on the n_samples/n_features regime.
Changes:
- Compute a data-dependent
alpha_maxfor pure ridge (l1_ratio=0) based on the null-model gradient (with a0.001surrogate divisor), instead of hardcodingalpha_max=10. - Change the default
min_alpha_ratioto1e-2whenn_samples < n_features(otherwise keep1e-6), matchingglmnet. - Update tests and documentation/changelog to reflect the new alpha-path behavior.
Reviewed changes
Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.
| File | Description |
|---|---|
src/glum/_glm.py |
Implements ridge alpha_max from null-model gradient; updates default min_alpha_ratio logic and docs. |
src/glum/_glm_cv.py |
Updates CV estimator docs for the new default min_alpha_ratio behavior. |
tests/glm/test_glm_regressor.py |
Extends alpha-path test coverage to include both lasso (l1_ratio=1) and ridge (l1_ratio=0). |
CHANGELOG.rst |
Adds an UNRELEASED entry describing the changed ridge alpha_max and min_alpha_ratio defaults. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
- Compute data-dependent alpha_max for pure ridge (l1_ratio=0) using the null model gradient divided by a small surrogate (0.001), instead of hardcoding alpha_max=10. Falls back to 10 when the gradient is zero (e.g. constant X after centering). - Default min_alpha_ratio is now 1e-2 when n_samples < n_features, matching glmnet's convention. Remains 1e-6 when n_samples >= n_features. - Extend test_alpha_path to cover the pure ridge case (l1_ratio=0). Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Compute data-dependent alpha_max for pure ridge (l1_ratio=0) using the null model gradient divided by a small surrogate (0.001), instead of hardcoding alpha_max=10. Falls back to 10 when the gradient is zero (e.g. constant X after centering). - Default min_alpha_ratio is now 1e-2 when n_samples < n_features, matching glmnet's convention. Remains 1e-6 when n_samples >= n_features. - Extend test_alpha_path to cover the pure ridge case (l1_ratio=0). Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Compute data-dependent alpha_max for pure ridge (l1_ratio=0) using the null model gradient divided by a small surrogate (0.001), instead of hardcoding alpha_max=10. Falls back to 10 when the gradient is zero (e.g. constant X after centering). - Default min_alpha_ratio is now 1e-2 when n_samples < n_features, matching glmnet's convention. Remains 1e-6 when n_samples >= n_features. - Extend test_alpha_path to cover the pure ridge case (l1_ratio=0). Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Copilot started reviewing on behalf of
MatthiasSchmidtblaicherQC
February 8, 2026 17:03
View session
There was a problem hiding this comment.
Pull request overview
Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
MatthiasSchmidtblaicherQC
requested review from
MarcAntoineSchmidtQC and
stanmart
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.
Closes #966.
The PR introduces a data dependent alpha grid for ridge regression (which we already have for lasso or elastic net) and aligns the alpha path computation more closely with glmnet:
Checklist
CHANGELOG.rstentry