NONLINEAR FUNCTIONS APPROXIMATION USING FAST WAVELET NEURAL NETWORK

In this paper we propose a fast learning algorithm of wavelet neural networks used in approximating nonlinear functions. Geometrical structure of the network is analyzed and the method of network parameters initialization is given. In the training phase, a new improved adaptive variable learning rate is proposed for fast convergence training. It is better in approaching the nonlinear functions than the traditional BP neural network. The results of simulation indicate that the method is fast in its convergence speed and has a good approaching precision. It provides a new method in modeling nonlinear systems besides offering a beneficial reference to the identification of complex nonlinear systems.