X
PrevPrev Go to previous topic
NextNext Go to next topic
Last Post 30 Mar 2016 02:57 PM by  anon
SVM method details
 1 Replies
Sort:
You are not authorized to post a reply.
Author Messages

anon



New Member


Posts:
New Member


--
30 Mar 2016 02:57 PM
    I have a few questions about the implementation of thesupervised classification method SVM. I'm trying to run grid search crossvalidation using python scikitlearn in order to identify the optimal C andgamma parameters for the RBF kernel. Here are my specific questions: 1) Can you confirm that SVM_DO_IT is based on the libsvm? Ifso, I should be able to use scikitlearn for cross validation since it is alsobased on libsvm. 2) Can you explain the data normalization procedure in moredetail? I understand that SVM_DO_IT uses a "linear normalization".What range are the values normalized to? [0,1] or [-1,1]? It's my understandingthat we need to scale training data used for cross validation in the samemanner as SVM_DO_IT so that the C and gamma parameters determined by crossvalidation can be used. 3) Assuming SVM_DO_IT is based on libsvm, are there anyother important differences I should be aware of as they pertain to crossvalidation?

    MariM



    Veteran Member


    Posts:2396
    Veteran Member


    --
    01 Apr 2016 02:38 PM
    From the Help help, it references the libsvm so it does use this: References Chang, C.-C. and C.-J. Lin. (2001). LIBSVM: a library for support vector machines. This help article might be useful for describing the normalization and cross-validation: http://www.harrisgeospatial.com/%20Su...
    You are not authorized to post a reply.