GPU accelerated libSVM grid search

This tool implements grid search of SVM parameters (cost and gamma),
based on NVIDIA CUDA framework, in svm-train. It takes advantages of
GPU parallel computing and support model level parallelization.

To run this tool, the CUDA toolkit and runtime environment should be
properly installed. This tool requires a minimum of 4GB GPU memory.



## Build ##
cd libsvm_cuda
make

## Build with NVML ##
make NVML_LIB=<nvml lib path>

Example: make NVML_LIB=/usr/lib/nvidia-352


## Run ##
Options: 
./svm-train-gpu -h

Example: find best combination of cost (C) and gamma (g):
./svm-train-gpu -C -5,15,2 -G 3,-15,-2 ./heart_scale 

Example: find cost (C) in SVM with linear kernel
./svm-train-gpu -t 0 -C -5,15,2 ./heart_scale 



Libsvm is a simple, easy-to-use, and efficient software for SVM
classification and regression. It solves C-SVM classification, nu-SVM
classification, one-class-SVM, epsilon-SVM regression, and nu-SVM
regression. It also provides an automatic model selection tool for
C-SVM classification. This document explains the use of libsvm.

Libsvm is available at 
http://www.csie.ntu.edu.tw/~cjlin/libsvm
Please read the COPYRIGHT file before using libsvm.


Author: Meng Wang <[email protected]>