A wavelet neural network for the approximation of nonlinear multivariable function

Abstract

Wavelet neural networks employing the wavelet function as the activation function have been proposed previously as an alternative approach to nonlinear mapping problems. In this paper, we propose a wavelet neural network which can be employed as a useful tool for learning a mapping between an input and an output space. The activation function of the proposed network is the compact supported non-orthogonal function which has been described by Yamakawa et al. (1996) as the convex wavelet in their paper. The proposed network can be proved to have the capability of approximating any continuous function in L/sup 2/. The experimental results of solving function approximation problems and a two-spirals classification problem indicate the better performance of the proposed network.