[pymvpa] SVM RBF kernel parameters

[Yaroslav Halchenko]

Hi Eric,

Unfortunately parameter selection is not yet a part of training.  We
have been cherishing an idea of OptimizedClassifier but haven't yet
implemented it yet. There is though ModelSelector which was
introduced by Emanuele to be used with GPR. For example of its usage
look into GPRWeights defined in gpr.py.

Sorry that pymvpa is not yet perfect ;)

Yarik

On Sun, 05 Apr 2009, Eric Trautmann wrote:

> I've looked through the API documentation but there doesn't seem to be  
> much info on parameter selection using the RBF kernel in a SVM  
> classifier.  The literature suggests various heuristic or brute-force  
> approaches for determining Gamma and C, though it's not clear to me if  
> this happens during classifier training:
>
> clf = RbfCSVMC()
>
> cv = CrossValidatedTransferError(TransferError(clf),  
> NFoldSplitter(cvtype=n), enable_states=['confusion', 'samples_error'])
>
> cv(dataset)
>
>
> When I run this code using default values of gamma or C, it generates a 
> result, but shouldn't this parameter selection be part of classifier  
> training?
>
> thanks,
> Eric
>
>
>
> _______________________________________________
> Pkg-ExpPsy-PyMVPA mailing list
> Pkg-ExpPsy-PyMVPA at lists.alioth.debian.org
> http://lists.alioth.debian.org/mailman/listinfo/pkg-exppsy-pymvpa
>

-- 
Yaroslav Halchenko
Research Assistant, Psychology Department, Rutgers-Newark
Student  Ph.D. @ CS Dept. NJIT
Office: (973) 353-1412 | FWD: 82823 | Fax: (973) 353-1171
        101 Warren Str, Smith Hall, Rm 4-105, Newark NJ 07102
WWW:     http://www.linkedin.com/in/yarik