I haven't actually run the code but I was looking through the linear RVM train and predict functions.
Imagining if the bias happens to be relevant the predict code first adds the ones to the trials and then selects features. Would this not discard the last relevant feature?
I think the one should be added after selecting the features. Maybe I am misunderstanding something?
I haven't actually run the code but I was looking through the linear RVM train and predict functions.
Imagining if the bias happens to be relevant the predict code first adds the ones to the trials and then selects features. Would this not discard the last relevant feature?
I think the one should be added after selecting the features. Maybe I am misunderstanding something?