Closed
Description
Describe the workflow you want to enable
Right now SVM supports gamma parameter to be set to 'scale' or 'auto', which will automatically determine gamma parameter based on data, but RBFSampler does not support this.
On top of that there is no check on n_components parameter, which allow users to set it to 0 and fit it without errors. It will only raise error if n_components is negative - ValueError: negative dimensions are not allowed (this is numpy error from random.normal generation, not sklearn one).
Describe your proposed solution
Add 'scale' and 'auto' option to RBFSampler
Describe alternatives you've considered, if relevant
It is possible to calculate gamma beforehand on your own, but it become increasingly tiresome if for example someone is performing CV - you will have to use custom loop instead of cross_val_score and others.
Additional context
No response