The Application of Dynamic K-means Clustering Algorithm in the Center Selection of RBF Neural Networks

Abstract

The key problem of constructing RBF neural networks is center selection. The method of adjusting the cluster centers is used in dynamic K-means clustering algorithm to make the choice of network-center more accurate. This paper, first introduced the structure of RBF Neural Networks (RBFNN) theory, and then applied the dynamic K-means clustering algorithm to the center selection of RBFNN. Our Simulation results show that the approximation of RBFNN, whose center selection is determined by the dynamic K-means clustering algorithm, has better performance and stronger practicality.