Initialization algorithm of wavelet networks and applications

Abstract

In this paper, the dilation and translation transforms of the mother wavelet are taken as the values at different points, and combining with the discreet wavelet transform theory and the mean idea, the mean model and related algorithms for one-dimensional (1-D) and a kind of multidimensional wavelet networks are proposed. And simulation instance analyses proved that the algorithm is effective, general and easy to be realized by programs. Performance comparisons of four different initialization algorithms are also proposed, and the comparison shows that the algorithm proposed in this paper is more effective and general than other algorithms.