FIX Fixes OneVsOneClassifier.predict for Estimators with only predict_proba #22604
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
|
Marking this as quick review. We do a very similar thing here: scikit-learn/sklearn/multiclass.py Lines 429 to 434 in 70eebb9 |
jjerphan
approved these changes
Mar 17, 2022
sklearn/multiclass.py
Outdated
Comment on lines
773
to
780
| if hasattr(self.estimators_[0], "decision_function") and is_classifier( | ||
| self.estimators_[0] | ||
| ): | ||
| thresh = 0 | ||
| else: | ||
| # predict_proba threshold | ||
| thresh = 0.5 | ||
| return self.classes_[(Y > thresh).astype(int)] |
There was a problem hiding this comment.
Is it worth wrapping this and the lines you mentioned, i.e.:
scikit-learn/sklearn/multiclass.py
Lines 429 to 434 in 70eebb9
in a private _threshold property?
I think this can make the logic explicit and ease maintainability.
There was a problem hiding this comment.
OneVsOneClassifier and OneVsRestClassifier do not share a common base class where a _threshold property makes sense. Also the threshold is only used in the binary case.
Inspired by your suggestion, I think a helper function makes the logic better 7372cee (#22604)
3290270 to
7372cee
Compare
|
Merging main to get rid of the doc-min-dependencies error about Python 3.7. |
|
Merging, thanks! |
glemaitre
pushed a commit
to glemaitre/scikit-learn
that referenced
this pull request
Apr 6, 2022
…_proba (scikit-learn#22604) Co-authored-by: Loïc Estève <loic.esteve@ymail.com>
joshua-cogliati-inl
added a commit
to joshua-cogliati-inl/raven
that referenced
this pull request
Dec 9, 2024
9 tasks
wangcj05
pushed a commit
to idaholab/raven
that referenced
this pull request
Jan 14, 2025
* Updating to scipy 1.12 * Scikitlearn 1.0 is incompatible with scipy 1.12, so updating to 1.1 * Sometimes only a 1d array is returned. * Removing testing kulsinski metric. * simps is deprecated so switching to simpson * Handle change from scikit-learn/scikit-learn#22199 * Improving test to classify better. * Regold due to scikit change scikit-learn/scikit-learn#22604 * Regold PolyExponential files (had rel err of 1e-03 or less) * Make timestep uniform for scipy update. * Regolding because of changes in scipy 1.12 * Increase limits to improve convergence. * Unpinning xarray and updating numpy * Updating various libraries. * Fix working with newer tensorflow. * Values need to be switched to tuples for hstack in numpy 1.26 * Updating to new ray version. * The deque size can be bigger in python 3.11 * Report difference in row lengths, instead of crashing OrderedCSVDiffer. Also report gold file name. * Remove Fourier__signal_f__period10.0__phase This was either +pi or -pi semirandomly, so nolonger testing it. * Regolding changes to ROM/TimeSeries/DMD/BOPDMD because of library changes. * Support xarray 2024.7 and newer. Pre 2024.7 automatically squeeze()ed groupby results, so now need to explicitly call squeeze(). * Fixing long line. * Increasing zero threshold because of change in libraries. * Remove version from setuptools since ray updated. * Optimizing persistence in BayesianMatyas. * Switch OVO to use estimator that is not constantly zero. * Use keepdims instead of try catch block. * Updating to default using python 3.11
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 #13617
What does this implement/fix? Explain your changes.
This PR uses the correct threshold if the inner estimator uses
predict_proba.