thundersvm on R caret

[asheetal]

Caret in R makes programming so easy and generates reusable code. I have benefited by embedding thundersvm in Caret. The problem I see (just waste some time) is that the model is not saved in the R object. I am not sure how thundersvm saved the R object. but when the session is exited, the object is lost as well. So in the new session, I just have to rebuild the thundersvm object from the saved hyperparameters, and that is works fine. Just adds 10 minutes of extra time on the dataset that I am using.

If thundersvm was added to caret it would be so useful. I love SVM that it quickly converges on GPU. using caret's framework