Implementation of k-Means Clustering Algorithm in CUDA
Dept. of Computer Science Engineering
International Journal of Enhanced Research in Management & Computer Applications, Vol. 3, Issue 9, pp. 15-24, 2014
@article{hooda2014implementation,
title={Implementation of k-Means Clustering Algorithm in CUDA},
author={Hooda, Hanu and Nandal, Rainu},
year={2014}
}
Big Data poses a very great computational challenge for programmers as well as machines as a lot of number crunching is to be done.Due to recent development in the shared memory inexpensive architecture like Graphics Processing Units (GPU), an alternative has emerged. In this paper, we target at decreasing runtime for k-Means, which is one of the most popular clustering algorithms, by using the widely available Graphics Processing Units (GPUs). The general – purpose applications are implemented on GPU using Compute Unified Device Architecture (CUDA). Cost effectiveness of the GPU and several features of CUDA like thread Divergence and coalescing memory access. Shared memory architecture is much more efficient than distributed memory architecture. Depending on hardware, data set, and k, dramatic improvements in performance can be seen over CPU implementations.
December 3, 2014 by hgpu
