2013/2/5 Gael Varoquaux <gael.varoqu...@normalesup.org>: > Do we need such genericity? Are there real practical gains to such > modularity? My gut feeling would be to move forward with something > simple, and make it performant statistical and computationaly for the > common cases.
Perhaps not, but would it hurt to have both? 2013/2/5 Olivier Grisel <olivier.gri...@ensta.org>: > Argl. This does not feel like linear model model at all. I would > rather put the transformer in sklearn.kernel_approximation . Indeed it doesn't :) +1 for kernel_approximations. However, if we do consider this a GLM, then that would warrant a coef_/intercept_ API on the transformer. @David: I just remembered that I already had a branch with an ELM classifier in it: https://github.com/larsmans/scikit-learn/tree/elm Maybe there's something in there that you could still use for yours? -- Lars Buitinck Scientific programmer, ILPS University of Amsterdam ------------------------------------------------------------------------------ Free Next-Gen Firewall Hardware Offer Buy your Sophos next-gen firewall before the end March 2013 and get the hardware for free! Learn more. http://p.sf.net/sfu/sophos-d2d-feb _______________________________________________ Scikit-learn-general mailing list Scikit-learn-general@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
