Kernel-based fuzzy C-means clustering algorithm for RBF network initialization

Abstract

Designing an effective structure of the RBF network is the task carried-out at the network initialization phase. Usual approach to deal with the problem is to decide on the number of hidden units and to apply a clustering algorithm to calculate cluster centroids. Clustering techniques have a strong influence on the performance of the RBF networks. The paper focuses on the radial basis function neural network initialization problem and the implementation of the kernel-based fuzzy C-means clustering algorithm, as an alternative method for the RBF networks initialization. Performance of the RBFNs initialized using the kernel-based fuzzy clustering algorithm is compared with several other clustering techniques, including k-means, fuzzy C-means and X-means.