Performance evaluation of Gaussian radial basis function network classifiers

There are various neural network techniques for pattern recognition and machine intelligence. Radial Basis Function Network (RBFN) has been shown as an important alternative to the conventional backpropagation approach in neural network design. A procedure to optimize the design parameters of the radial basis function classifier is described here. We are evaluating the results of the standard radial basis function classifier, its optimized version and the backpropagation classifier in terms of training speed and classifier accuracy. An artificial two-dimensional data set is created for our study.