Batch implementation of Growing Self-Organizing Map

The growing self-organizing map algorithm (GSOM) has been developed based on SOM. With its dynamic structure, GSOM can generate feature maps without pre-determining their size. It has shown many significant advantages when processing very large data sets. These advantages could be further enhanced if the processing speed of the algorithm could be increased. This paper presents a modified version of GSOM, which implements the batch processing principle to shorten the processing time. The algorithm and experimental results showing the improved performance are presented in this paper.