LCSL / GURLS

GURLS: a Least Squares Library for Supervised Learning
http://lcsl.mit.edu/#/downloads/gurls
63 stars 37 forks source link

Random features for Gaussian kernel approximation functions provide no sigma parameter explicit choice, nor associated paramsel method #20

Open raffaello-camoriano opened 9 years ago

raffaello-camoriano commented 9 years ago

When working with random Fourier features (RFs) for the approximation of a Gaussian kernel, it would be very useful to have the following functionalities: