MODNet model without feature_selection()

[mhlee216]

How can I use MODNet model without md.feature_selection(n=-1)?

model.fit() doesn't work because of this error.

AttributeError: 'MODData' object has no attribute 'optimal_features'

md.get_optimal_descriptors = df_with_my_features didn't solve it either.

I want to use my DataFrame(csv file) as MODData.

[ml-evs]

Hi @mhlee216, if I am understanding your problem correctly, I think you can simply set md.optimal_features (not get_optimal_descriptors) with the column names of the descriptors you want to use within your dataset. If you are doing multi-target learning you may also need to set md.optimal_features_by_target to a dictionary with structure {target: [md.optimal_features]} for target in targets}.

We could certainly provide a nicer interface for this workflow, but the feature selection aspect is a big part of the MODNet approach.

[mhlee216]

Thank you!