In this paper we introduce a MapReduce-based implementation of self-organizing maps that performs compute-bound operations on distributed GPUs. The kernels are optimized to ensure coalesced memory access and effective use of shared memory. We have performed extensive tests of our algorithms on a cluster of eight nodes with two NVidia Tesla M2050 attached to each, and we achieve a 10x speedup fo...

This paper deals with the Kohonen Self-organizing maps for cluster analysis applications. The cluster analysis represents a group of methods whose aim is to classify the objects into clusters. For solving cluster analysis applications many new algorithms using neural networks have been used. This paper describes the use of an advanced method of neural network represented by Kohonen self-organiz...