ENH add new common test: test_negative_sample_weight_support and allow_negative_sample_weight tag
#21132
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
…ase error was wrong and some tag changes where it was needed
…e it fails tests and previously it wasnt there.
… it wasnt there before.
|
I wonder if we should assume that negative weight work by default. I don't really see a valid use case for negative weights and don't want to enforce complexity of handling this edge case if it's not any practical use. |
|
Related: #15657. |
Good point, Maybe the default |
|
Closing as there hasn't been further activity from the OP. |
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Outdated suggestions cannot be applied.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 #15531.
What does this implement/fix? Explain your changes.
A new tag is added:
allow_negative_sample_weight. By default it is True, and it is changed to False if the estimator doesntsupports negative
sample_weight. In this case there is a check for the negative case. This far the following estimatorschecked for negative weights:
LinearRegression,AdaBoostClassifier,AdaBoostRegressor, andKDEDensity.Some estimators raised misleading error in the negative weight case:
BayesianRidge,KernelRidge,Ridge,RidgeClassifier,ElasticNetCVandLassoCV. These are updated. Some estimators which uses another estimator copies its allow_negative_sample_weight tag. It is needed because those estimators fit method will be used, so they have to handle sample_weights the same way:RANSACRegressor,StackingRegressor,StackingClassifier,VotingRegressor,VotingClassifierandMultiOutputEstimator.A common test is added:
test_negative_sample_weight_support. This checks all estimators in_tested_estimators()where the estimator hassample_weightparameter infit. If the estimator has False tag then the appropriate error is expected, else no error.A helper function is added:
_create_data_for_fit. This function is needed to run thefitmethod in the test.Any other comments?
Now every estimator in
_tested_estimators()is explicitly tested by using neagative weights in fit and check for result. There is no estimator which usessample_weightinfitand it is not listed in_tested_estimators(). The changes are backward compatible, because only those cases were updated where the negative weight case raised another error so far.