Describe the issue linked to the documentation

Hi folks,

I am somewhat in doubt that the sklearn.feature_selection.chi2 function is implemented correctly. At least, checking the source code, it is entirely unclear to me why that kind of scoring would make sense.

Best, Felix

Suggest a potential alternative/fix

Is there any reference that justifies the computation in the implemented way? Then it should be added to the documentation.

© 2022 - All rights reserved.