GPU-accelerated LIBSVM

A GPU-accelerated implementation of LIBSVM was released. This is a modification of the original LIBSVM that exploits the CUDA framework to significantly reduce processing time while producing identical results. GPU-accelerated LIBSVM gives a performance gain depending on the size of the input data set; this gain increases significantly when considering large data sets. More details and the software can be downloaded from http://mklab.iti.gr/project/GPU-LIBSVM