Currently, MutualInformationFeatureSelector cannot be incorporated into a sklearn.Pipeline, because that expects the class to have a function set_params to be present for setting the hyper parameters (which are now passed at instantiation)
By inheriting from the sklearn base classes, this functionality is added, while at the same time allowing for a simplification of the code (less functions need to be defined).
Furthermore, removed the search for k (replaced by user defined k)
danielhomola
commented
5 years ago
Currently, MutualInformationFeatureSelector cannot be incorporated into a
sklearn.Pipeline, because that expects the class to have a functionset_paramsto be present for setting the hyper parameters (which are now passed at instantiation)By inheriting from the sklearn base classes, this functionality is added, while at the same time allowing for a simplification of the code (less functions need to be defined).
Furthermore, removed the search for
k(replaced by user definedk)