Revision history [back]

you can use:

 svm->predict(vdata, scores, ml::StatModel::RAW_OUTPUT);

note, that this is the distance to the margin, not a probability.