[MRG] Fix label-aware cost correction in ot.da (closes #664)

[CodingSelim]

Types of changes

• Bug fix (non-breaking change which fixes an issue)

Motivation and context / Related issue

Closes#664.

The cost correction in BaseTransport.fit that is supposed to push apart source and target samples with different labels was computed backwards. It built missing_ys = (ys == -1) / missing_yt = (yt == -1) and used their outer product missing_labels, so the large limit_max cost was applied only where both labels are missing, and label_match was actually a mismatch mask. The net effect:

• when all labels are known (missing_labels is all zeros) the correction is a complete no-op, so ys/yt had no influence on the transport at all
• it also contradicted the code`s own comment ("0 iff either masked").

This is why a supervised fit gave the same cost matrix as an unsupervised one.

Fix: use present masks (ys != -1) / (yt != -1), so the correction applies exactly to labeled source/target pairs whose labels differ. This matches the original pre-vectorized loop (the for c in classes: cost_[idx_s, idx_t] = limit_max version) that the vectorization in #587 replaced.

How has this been tested

• The semisupervised checks in test_da.py had been flipped when the bug was introduced to assert n_unsup == n_semisup (i.e. that supervised mode does nothing), while the comment right above still said the norms should be different. Restored them to assert the cost actually changes.
• Added test_semisupervised_cost_correction: with all labels known, different-label pairs get pushed to limit_max and same-label pairs keep their smaller cost; in the semisupervised case, columns with an unlabeled (-1) target are never forbidden. It fails on current master and passes with this change.
• Full test_da.py passes locally.

PR checklist

• I have read the CONTRIBUTING document.
• All tests passed, and additional code has been covered with new tests.
• I have added the PR and Issue fix to the RELEASES.md file.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 96.85%. Comparing base (e52c2a3) to head (fbd765f).

Additional details and impacted files

@@ Coverage Diff @@## master #833 +/- ##
=======================================
Coverage 96.85% 96.85% =======================================
Files 128 128 Lines 25659 25683 +24 =======================================
+ Hits 24852 24876 +24 
Misses 807 807 

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[rflamary]

hello @CodingSelim the PR is looking good, what you coded makes sens to me. You need to skip the tf backend for the DA class.

@kachayev i'm open to your opinion on this too. I would like to do a release shortly

[kachayev]

@rflamary Looks good to me! Left a couple of minor comments/suggestions.

[CodingSelim]

thanks for the review! pushed the MISSING_LABEL constant.

on the cost_correction refactor, i think the item assignment (cost_correction[mask] = self.limit_max) won't work across backends, jax arrays are immutable and the tf tensor path doesn't support that kind of in-place scatter either, which is why the original avoids indexing and does it with arithmetic. i kept the multiply approach but it should be fine, the only reason for the catch_warnings is the 0*inf case when limit_max is left at the default inf, and we already scrub those with nan_to_num right after. happy to try a where-based version if you'd prefer that over the warnings filter, that one stays backend safe.

[rflamary]

Yep jax is a pain for generic backends... I think the PR looks great, let us wait for tetss to pass and merge it

[kachayev]

@CodingSelim you are right! don't worry, it's a chunk of old code anyways