I'm using a SVM classifier with a weighted RBF kernel. My dataset has 17 features. In the RBF kernel I will use a weight for each feature. Of course the weights must sum to one. For choosing the best weights, I'm using random search.

Should I you use in this case the stick-breaking process or the Dirichlet pseudo random number generator?

Second, should I vary the weights only around 1/17 (e.g. with a normal distribution) or should I allow some few weights to become very large and the others very small.



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.