scikit-learn-contrib / metric-learn

Metric learning algorithms in Python
http://contrib.scikit-learn.org/metric-learn/
MIT License
1.4k stars 234 forks source link

separate list of mahanobis metric learners and general metric learners in tests #160

Open wdevazelhes opened 5 years ago

wdevazelhes commented 5 years ago

Currently a lot of tests use a list of metric learners and their dataset, defined at the top of test_utils. The problem is that there is no difference currently between a list of all metric learners and a list of mahalanobis metric learners. This is fine for now because there are no not Mahalanobis metric learners, but it will be problematic when we have one, because then some tests that are made only for Mahalanobis metric learners should be ran only on such metric learners. This will force to look through the tests to separate those for mahalanobis and those that are not for mahalanobis. Right now, tests for mahalanobis metric learners are in test_mahalanobis_mixin, so it will not be difficult eventually, but I think it's good to do the change now that there are not so much tests.

bellet commented 5 years ago

This is related to #136