How to perform multi-class cross-validation for LIBSVM in MatLab -
i want use libsvm in matlab multi-class classification. have read libsvm use 1 vs. 1 default when provided multiple labels, , fine it.
my question parameter search , model validation. when doing 2-class validation find parameters c , gamma (when using rbf kernel), use built-in cross validation find best (c,gamma)-pair, using simple grid search. have read libsvm documentation have no idea how validation works multiclass svm.
does built-in option returns multi-class accuracy? how can provide best parameters each of ovo models automaticaly built?
the answer given there http://www.csie.ntu.edu.tw/~cjlin/libsvm/faq.html#f507. did not read faq of libsvm enough.
Comments
Post a Comment