GUCAS_CUMiner

Description

GUCAS_CUMiner is a suite of CUDA-based data mining tools. This suite is intended to speed up several data mining algorithms by utilitying NVIDIA's GPU and parallel computing architecture-CUDA.

With the rapid development of GPU, tremendous memory bandwidth and computing power are available and have been used to enhance traditional general scientific computing in many other research areas. For data mining applications, we select several typical data mining algorithms, which are computational-intensive, and implement them on NVIDIA GPU's architecture.
 

We adopted the strategy of "do what you are good at" to make both CPU and GPU work at full speed.

Acknowledgements

This project is partially supported by the NVIDIA’s global Professor Partnership, 2009.

Publications

Liheng Jian, Ying Liu, Cheng Wang, Peng Zhang, “CUApriori: A Parallel Implementation of Association Rules Mining on CUDA-enabled GPU", Technical Report, #GUCAS-2009-08-01, Graduate University of Chinese Academy of Sciences.

Shenshen Liang, Ying Liu, Cheng Wang and Liheng Jian, “A CUDA-based Parallel Implementation of K-Nearest Neighbor Algorithm”, International Conference on Cyber-Enabled Distributed Computing and Knowledge Discovery (CyberC), October 2009.

Download

GUCAS_CUMiner-1.1 for linux platform

PLEASE NOTE: GUCAS_CUMiner is a copyright of Ying Liu's High Performance Data Mining Research Group at Graduate University of Chinese Academy of Sciences. At the present, the executables of three algorithms are available. All rights reserved.

Contact Information

In case you have irresolvable issues or if you would like to give suggestions or contribute software, please email us at yingliu@gucas.ac.cn.