Abstract Views :190 |
PDF Views:2
Authors
Affiliations
1 Mumbai, Maharashtra, IN
2 Oracle Financial Services Software, Mumbai, Maharashtra, IN
3 IIM, Lucknow, Uttar Pradesh, IN
4 Department of Electrical and Computer Engineering, Carnegie Mellon University, US
Source
International Journal of Innovative Research and Development, Vol 5, No 14 (2016), Pagination: 55-61
Abstract
This paper focuses on an innovative approach of implementing parallel processing using NVIDIA’s Graphics Processing Unit (GPU) to accelerate a data mining application in R. In order to accomplish this, one of the most apposite and efficient solution is to use CUDA (Compute Unified Device Architecture). We have used the k-means clustering algorithm to demonstrate the effectiveness of CUDA C in terms of speed-up and reduced latency. It is a widely used unsupervised learning technique in data science applications. The currently existing sequential R programming technique using C for k-means algorithm was converted to a more optimized and efficient code that uses concepts of parallel computing using GPU. The efficiency of C and CUDA C codes has been compared on the basis of execution time.
Keywords
CUDA, NVIDIA, GPU, Parallel Processing, R, CUDA C, Clustering, K-Means Algorithm, Llyod Algorithm, Data Mining.
Full Text
