FIX: Fix QuantileTransformer(ignore_implicit_zeros=True) sub-sampling behavior
#32587
+51
−5
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
cakedev0
changed the title
QuantileTransformer(ignore_implicit_zeros=True) sub-sampling behaviorQuantileTransformer(ignore_implicit_zeros=True) sub-sampling behavior
|
I'd add a change log entry, even if no one has reported a bug people might notice a change in behaviour. The entry is the way we want people to answer that kind of question for themselves without posting here. Fix looks good to me. Not thought about the test yet. |
mdshoaibuddinchanda
pushed a commit
to mdshoaibuddinchanda/scikit-learn
that referenced
this pull request
Nov 2, 2025
…bug (scikit-learn#32587) Add deterministic regression test that constructs a very sparse CSC matrix with two nearly-identical columns (one with subsample - 1 nonzeros, the other with subsample + 1 nonzeros) and verifies that the transformed columns are not degenerate after fit/transform. For large declared n_samples, the bug could create a zero-sized column sample leading to collapsed quantiles (all zeros) for some columns. References: scikit-learn#32587
mdshoaibuddinchanda
pushed a commit
to mdshoaibuddinchanda/scikit-learn
that referenced
this pull request
Nov 2, 2025
…eros=True When ignore_implicit_zeros=True, use subsample directly instead of scaling by n_samples to avoid zero-sized column samples. Also add max(1) guard for the else case. Fixes scikit-learn#32587
mdshoaibuddinchanda
pushed a commit
to mdshoaibuddinchanda/scikit-learn
that referenced
this pull request
Nov 2, 2025
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 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Outdated suggestions cannot be applied.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.
Reference Issues/PRs
Fixes: #32585
What does this implement/fix? Explain your changes.
Fix the logic for the sub-sampling (only affect the case
ignore_implicit_zeros=True)Adds a non-regression test.
Any other comments?
While reviewing the PR, you might notice "behavior discontinuity" that this PR fixes also exists for the case
ignore_implicit_zeros=False. But for this case, it doesn't incur unexpected/incorrect results, just unexpectedly long/short runtime. See issue [TODO] about that.I haven't written a change-log as nobody reported the bug except me, and I found it just by reading the code... Let me know if this is needed.