Ensemble of Radial Basis Neural Networks with Chinese Restaurant Process

Abstract

Ensemble classifier implementation needs several considerations including base classifier selection and decision aggregation. A set of radial basis function networks is one of the most popular method as a base classifier. However, considering that the unsupervised method including clustering is frequently applied in the learning schemes of the radial basis function networks, there is an important issue to solve that the number of cluster must be determined in advance. Most of partitional clustering algorithms including k-means clustering are sensitive to the number of clusters. In this paper, we replace the k-means clustering algorithm in the learning scheme into the Chinese restaurant process, which does not need to determine the number of cluster. With real problems in the radar data analysis, the proposed method shows better results in the experiment.