Combining RBF Networks Trained by Different Clustering Techniques

Abstract

Clustering techniques have a strong influence on the performance achieved by Radial Basis Function (RBF) networks. This article compares the performance achieved by RBF networks using seven different clustering techniques. For such, different sizes of RBF networks are trained and tested using an Automatic Target Recognition data set. The performances of these RBF networks using each clustering technique are compared and analyzed. This article also evaluates how the performance can be improved by combining RBF networks, training with different clustering techniques, in committees.