Identification of radial basis function networks by using revised GMDH-type neural networks with a feedback loop

Radial basis function networks are identified by using a revised GMDH-type neural network with a feedback loop. Conventional radial basis function networks have one hidden layer which makes their architecture simple. Nevertheless, it is difficult to learn the parameters of the hidden layer. Therefore, good approximation of very complex nonlinear systems cannot be achieved by using the conventional radial basis function networks. The revised GMDH-type neural networks with a feedback loop proposed in the paper can identify the radial basis function networks accurately because the complexity of the neural networks is increased gradually by the feedback loop calculations. Furthermore, the structural parameters such as the number of the neurons, useful input variables and the number of feedback loop calculations are automatically determined so as to minimize the prediction error criterion defined as AIC.