A Dual Structural Radial Basis Function Network for Recursive Function Estimation∗

We present a dual structural radial basis function (RBF) network for recursive function estimation. This network is a hybrid system which consists of two sub-RBF networks. One sub-network models the relationship between the current network output and the past ones, and the other one describes the relationship between the current network output and the inputs. We propose a new variant of extended normalized RBF (ENRBF) network to implement each subRBF net. This variant uses an adjustable p-order single-term polynomial, rather than the first-order one, to fit the relations between each hidden unit and each output unit. It not only includes the existing ENRBF net as its special case, but also has better fitting ability in general under the moderate number of hidden units. The experiments have shown the proposed net’s performance.