FIX: Repair PCA array API support tag for svd_solver "full", "covariance_eigh", or "auto" #32198
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
OmarManzoor
reviewed
Sep 16, 2025
Contributor
OmarManzoor
left a comment
There was a problem hiding this comment.
Thanks for the PR @icfaust
If the covariance_eigh solver supports the array api then you can simply add in the current docs for array_api where PCA is already mentioned and add covariance_eigh as an additional value.
OmarManzoor
approved these changes
Sep 17, 2025
Contributor
OmarManzoor
left a comment
There was a problem hiding this comment.
LGTM. Thanks @icfaust
Contributor
9 tasks
lesteve
reviewed
Sep 22, 2025
| self.svd_solver in ["full", "randomized"] | ||
| and self.power_iteration_normalizer == "QR" | ||
| solver = getattr(self, "_fit_svd_solver", self.svd_solver) | ||
| tags.array_api_support = solver not in ["arpack", "randomized"] or ( |
Member
There was a problem hiding this comment.
Thanks for fixing this!
I would keep a whitelist here because for now array API is available in a few cases. It also makes it easier to check that the documentation and the code match.
If the documentation is correct I would expect something like this?
tags.array_api_support = solver in ["full", "covariance_eigh"] or (solver == "randomized" and self.power_iteration_normalizer == "QR")
Member
|
Merging this one, thanks! |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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Outdated suggestions cannot be applied.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 on to #31784 (which is a follow-up to #30777), the conditions laid out in documentation for PCA: https://scikit-learn.org/dev/modules/array_api.html#estimators states:
The documentation for PCA states for
power_iteration_normalizer("https://scikit-learn.org/stable/modules/generated/sklearn.decomposition.PCA.html) that:This is leading to array API support failures in our repository (uxlfoundation/scikit-learn-intelex#2578) as we prep for your 1.8 release.
The logic should swap from a whitelist to a blacklist to speed and simplify logic, thereby only checking the power_iteration_normalizer if svd_solver is either 1) "randomized" or _fit_svd_solver is set to "randomized". This corrects the logic to probe for the
_fit_svd_solverif available (to handle thesvd_sovler=autocondition) or fall back to checkingsvd_solverif the PCA estimator has not been fit.I'm not sure how to update the documentation to reflect this as the 'covariance_eigh' solver also supports array API but is not reflected in the documentation.
@OmarManzoor @lesteve