In scikit-learn, some estimators support class_weight and sample_weight.

It might be worth testing the correct interaction of those two types of weights, especially asserting that:

• setting a class weights to zero is equivalent to excluding the samples associated to this class from the calibration even when using non uniform sample weights.
• setting some samples weights to zero is equivalent to excluding those samples from the calibration even when if they are associated to using non uniform class weights.

Relevant interfaces:

• the main subclasses of sklearn.tree.BaseDecisionTree for classification, i.e.:
  • sklearn.tree.DecisionTreeClassifier
  • sklearn.tree.ExtraTreeClassifier
• the main subclasses of sklearn.ensemble.BaseForest for classification and embedding, i.e.:
  • sklearn.ensemble.RandomTreesEmbedding
  • sklearn.ensemble.RandomForestClassifier
  • sklearn.ensemble.ExtraTreesClassifier
• sklearn.linear_model.LogisticRegression
• sklearn.linear_model.LogisticRegressionCV
• sklearn.CalibratedClassifierCV after the merge of [MRG] Add class_weight parameter to CalibratedClassifierCV #17541