Radial Basis Function Neural Networks : A Review

Radial Basis Function neural networks (RBFNNs) represent an attractive alternative to other neural network models. One reason is that they form a unifying link between function approximation, regularization, noisy interpolation, classification and density estimation. It is also the case that training RBF neural networks is faster than training multi-layer perceptron networks. RBFNN learning is usually split into an unsupervised part, where center and widths of the Gaussian basis functions are set, and a linear supervised part for weight computation. This paper reviews various learning methods for determining centers, widths, and synaptic weights of RBFNN. In addition, we will point to some applications of RBFNN in various fields. In the end, we name software that can be used for implementing RBFNNs.