Do we want to validate data also for score method of the EmpiricalCovariance class?

The common test does not currently cover score. We most likely need to go through all the modules to add _validate_data to score.

But in this specific case, EmpiricalCovariance.score calls predict, it will get validated.

Edit: I looked at the wrong score method.

But in this specific case, EmpiricalCovariance.score calls predict, it will get validated.

Shouldn't there be a self._validate_data in def score and def mahalanobis of class EmpiricalCovariance?

Shouldn't there be a self._validate_data in def score and def mahalanobis of class EmpiricalCovariance?

score makes sense, as for mahalanobis as long as we use config_context(assume_finite=True) to turn off the nan check. Because pairwise_distances calls check_pairwise_arrays, which would validate again.

I think score should be a part of the common test, and mahalaobis needs its own test.

Looking at the code base, adding score to the common test should not be too difficult because most of it comes from the mixin calling predict, where predict already validates.

Updated PR with adding score to the common test and a test for mahalanobis.

Thank you @thomasjpfan