The expansibility research of K-Means algorithm under the GPU

Abstract

K-Means algorithm is one of the most popular clustering analysis algorithm. Since the algorithm can be easily understood and implemented, and its execution is more efficient than common clustering algorithm, it has been used widely. At the same time, with the increasing size of the data sets processed, CPU-based serial K-Means implementation has been unable to meet the people's needof data processing. Parallel computing is considered well with the large data sets tasks. GPU-based concurrent computation can accelerate common tasks, and especially for accelerating the compute-intensive tasks. CUDA (Compute Unified Device Architecture) is one of the methods that achieving the GPU-based concurrent computation. In the paper, the author hope to achieve a K-Means algorithm implementation can handle larger data sets via CUDA and the algorithm can be used on a common computer with NVIDIA graphics cards.