On 03/17/2012 12:40 AM, Olivier Grisel wrote: > Le 16 mars 2012 15:29, Andreas Müller<[email protected]> a écrit : > >> On 03/16/2012 11:12 PM, James Bergstra wrote: >> >>> I was also recently bit by this scale_C business. It looks like the >>> decision to make scale_C=True the un-changeable default has already >>> been made, but when this is done *PLEASE* make this abundantly clear >>> in the documentation... my understanding was that the C in the SVM >>> equation means the thing that Ian referred to, and which libSVM uses, >>> and shogun, etc... rather than this "conveniently scaled" >>> alternative. Maybe this new parameter should be renamed to something >>> other than "C" so that SVC(C=foo) fails rather than computes a >>> surprising thing? Anyway, I'm sure this is off-topic, but if someone >>> could point me to a more appropriate discussion thread I'd appreciate >>> it. >>> >> >> Hi James. >> There was some discussion on this going on pretty invisible here: >> https://github.com/scikit-learn/scikit-learn/commit/b3839837cf949db8e9122807578ebce6bf3d6fb4#commitcomment-1022876 >> >> Maybe we should open an issue on this? >> Alex wanted to change the default behavior for "C". >> I'm not sure if this will actually make everything ok again ;) >> (Also I'm not sure if this was maybe already merged?) >> >> I have to say that I am also not entirely happy with this step >> but changing the API _again_ would be ... well... pretty >> unfortunate. >> > AFAIK the new "scale_C=true" by default has not been released yet. > > I agree, that using C for a scaled regularizer parameter is misleading > for people who are familiar with the libsvm and liblinear convention. > That's why I proposed to rename it use "alpha" instead of C to be > consistent with the other regularized models of the project (and esp. > linear models such as RidgeRegression, SGDClassifier, ElasticNet, > LassoLars and friends). > > If we change the API, I would go for alpha as the current "scale_C=True" but optionally provide the "C", which behaves like the LibSVM parameter.
------------------------------------------------------------------------------ This SF email is sponsosred by: Try Windows Azure free for 90 days Click Here http://p.sf.net/sfu/sfd2d-msazure _______________________________________________ Scikit-learn-general mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
