GPUML GPUs for kernel machines 4

Balaji Vasan Srinivasan, Qi Hu, Ramani Duraiswami — Fri, 26 Feb 2010 18:12:46 -0000

Kernel machines are integral part of many learning approaches. We have identified 3 key computations encountered in these algorithms. 1. Weighted summation of kernel function (ex. kernel density estimation) 2. Kernel matrix-vector product in an iterative algorithm (ex. kernel regression) 3. Operations on the kernel matrix (ex. kernel PCA)

In this software, we accelerate each of these on a GPU. See the documentation for more details, and demo to see the speedups.

System requirement: Cuda 2.2+, Visual studio 2008 (if used in Windows)

mloss.org GPUML GPUs for kernel machines

GPUML GPUs for kernel machines 4