Ensemble of structure-adaptive self-organizing maps for high performance classification

Combining multiple models has been recently exploited for the development o f reliable neural networks. This paper introduces a structure-adaptive self -organizing map (SOM) which can adapt the structure as well as the weights, and presents a method to improve the performance by combining the multiple maps. The structure-adaptive SOM places the nodes of prototype vectors int o the pattern space properly so as to make the decision boundaries as close to the class boundaries as possible. In order to show the performance of t he proposed method, experiments with the unconstrained handwritten digit da tabase of Concordia University in Canada have been conducted, (C) 2000 Else vier Science Inc. All rights reserved.