Getting SVM weights from svmtrain (using sigmoid kernel)

3 views (last 30 days)
I've trained an SVM on my data using svtrain and a sigmoid kernel. After the fact, I'd like to visualize how the weights are distributed across features, so that I can see how much each feature contributes to the model, however I can't find anything within the SVM struct that gives me something like this. In a linear SVM I'd have the "w" from "w.x-b" however I'm not certain how this works with kernels.
  1 Comment
Dail
Dail on 30 Jul 2014
I have the same question as u do. I tried to use (model.Alpha.*tY(model.SupportVectorIndices))'*model.SupportVectors to get w, but it doesn't match the plotted data. Besides, model is a SVMSTRUCT object.

Sign in to comment.

Answers (0)

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!