Best MATLAB toolbox that implements Support Vector Regression?

Question

In this Wikipedia article about SVM there are a number of links to different implementations of MATLAB toolboxes for Support Vector Machines. Could anyone suggest which of these is best in terms of speed, ease of use, etc.?

Answer 1

I've used libSVM. It is pretty fast and easy, and provides some useful tools, too. There are some examples of it in use here. The other nice thing is that there are implementations in C++ and Java, too, so if you find yourself needing to develop outside of Matlab (to turn a prototype into something speedy, for example), you will have a familiar interface to work with.

Answer 2

Without a doubt Cawley's is the best.

http://theoval.sys.uea.ac.uk/svm/toolbox/

Answer 3

You can always use MathWorks implementation of SVM in the Bioinformatics Toolbox with the functions: svmtrain and svmclassify, which as usual have excellent documentation

Answer 4

This doesn't answer your question directly, but if you want to speed up an M script look into the embedded matlab toolbox and MEX functions. Basically, you can use these tools to compile your M scripts, I've done it and I get an order of magnitude performance gain minimum. The people at the MW say you can get 100 fold improvement.