Wavelet neural network for function approximation and network optimization
Kunikazu Kobayashi, Toyoshi Torioka · 1994
: A new mapping network combined wavelet and neural networks is proposed. The algorithm consists of two process: the selfconstruction of networks and the minimization of errors. In the rst process, the network structure is determined by using wavelet analysis. In the second process, the approximation errors are minimized. The merits of the proposed network are as follows: network optimization, partial retrieval of the approximated function, fast convergence and escaping local minima. The computer simulations conrmed these merits INTRODUCTION Recently, it has been shown that neural networks (NNs) can realize any mappings (e.g., Hecht-Nielsen, 1987). These are important to theoretically explore the potential of NNs not practically. Backpropagation (BP) networks are now the most popular mapping network (Rumelhart, Hinton and Williams, 1985). It is, however, well known that BP networks have few problems such as trapping into local minima and slow convergence. In addition, the ...