Chaotic time series prediction by combining echo-state networks and radial basis function networks

In this paper, we describe a chaotic time series prediction using a combination of an echo state network (ESN) and a radial basis function network (RBFN). The ESN is a neural network consisting of three layers, where the hidden layer (the “reservoir”) is composed of many neurons. The RBFN is a neural network using a radial basis function (RBF) for its output function. We propose a neural network model which is a combination of the ESN and the RBFN. Time series predictions for the Mackey-Glass equation of a chaotic time series and the laser time series are examined. Numerical experiments to examine the efficiency of the proposed network model reveal that the proposed combined model shows higher prediction ability than the conventional ESN model.