Fix for Multiclass SVC.fit fails if sample_weight zeros out a class #26593
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
…s SVM with a new warning, added test
adrinjalali
reviewed
Jun 22, 2023
There was a problem hiding this comment.
So in terms of API, regardless of whether a class is zeroed out, it should be included in classes_, the same way it is for other cases (dense, LinearSVC, etc)
To be consistent, the fix needs to be in svm.cpp, the same way it's done for the other case, which you see around this line:
scikit-learn/sklearn/svm/src/libsvm/svm.cpp
Line 2465 in 04575f5
It's not gonna be easy 😁
| @@ -144,7 +143,6 @@ def _more_tags(self): | |||
| # Used by cross_val_score. | |||
| return {"pairwise": self.kernel == "precomputed"} | |||
|
|
|||
| @_fit_context(prefer_skip_nested_validation=True) | |||
There was a problem hiding this comment.
I think this is an unintended change, should be put back
Comment on lines
+179
to
+180
| self._validate_params() | ||
|
|
There was a problem hiding this comment.
same here, should be removed
|
Closing this PR because it does not seem to be the right fix: #25380 (comment) |
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Outdated suggestions cannot be applied.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.
Reference Issues/PRs
Fixes #25380
What does this implement/fix? Explain your changes.
Fix for multiclass SVM when weights of a class are zero
Now it removes the class with zero weights from X, y, and sample_weight
Trains the model and skips zero classes
Also, a test to assert model params when it is trained with or without the class
A new warning for removing the weights has been added
Any other comments?
The labels of y remain unchanged