TR + Code for automatic C, sigma tuning in SVMs

[S. Sathiya Keerthi]

Efficient Tuning of SVM Hyperparameters Using Radius/Margin Bound 
and Iterative Algorithms

S. Keerthi

National Univeristy of Singapore
mpessk at guppy.mpe.nus.edu.sg

Motivated by a recent SVM paper by Chapelle, Vapnik, Bousquet and
Mukherjee, I have been working on a code for the automatic tuning
of the hyperparameters of a Support Vector Machine with L2 soft
margin, for which the radius/margin bound is taken as the index to be
minimized and iterative techniques are employed for computing radius
and margin. The implementation is found to be feasible and efficient
even for large problems having more than 10,000 support vectors. A
report discussing the implementation issues, together with a code
running on Matlab interface, can be downloaded from the following page:

http://guppy.mpe.nus.edu.sg/~mpessk/nparm.shtml 

Any comments on the tech report as well as the performance of the code
on new datasets will be very much appreciated.