An on-line learning algorithm with dimension selection using minimal hyper basis function networks

Abstract

In this study, we extend a minimal resource-allocating network (MRAN) which is an on-line learning system for Gaussian radial basis function networks (GRBFs) with growing and pruning strategies so as to realize dimension selection and low computational complexity. We demonstrate that the proposed algorithm outperforms conventional algorithms in terms of both accuracy and computational complexity via some experiments. © Springer-Verlag Berlin Heidelberg 2004.