The Effect of Random Weight Updation in Dynamic Self Organizing Maps

The random weight adaptation scheme presented in this paper is capable of simulating the effect of presenting the inputs in a random order to self-organizing map algorithms. The resulting effect enables the inputs to be presented in sequential order and still achieve results similar to that of presenting the inputs in random order. This capability enables efficient processing of massive datasets. The random weight adaptation is implemented on a growing variant of the self organizing map algorithm called the high dimensional growing self organizing map (HDGSOM) to demonstrate the efficiency of the new weight adaptation scheme. Several experimental results using this new algorithm are also presented.