high performance computing on graphics processing units: hgpu.org

ISODATA is a well-known clustering algorithm based on the nearest neighbor rule, which has been widely used in various areas. It employs a heuristic strategy allowing the clusters to split and merge as appropriate. However, since the volume of the data to be clustered in the real world is growing continuously, the efficiency of the serial ISODATA has become a serious practical issue. The GPU (Graphics Processing Unit) is an emerging high performance computing platform due to its highly parallel multithreaded architecture. In this paper, we propose an efficient parallel ISODATA algorithm based on the latest Kepler GPUs and the dynamic parallelism feature in CUDA (Compute Unified Device Architecture). Performance study shows that our parallel ISODATA can achieve promising speedup ratios and features favorable scalability compared to the original algorithm.