小波变换与小波神经元的选择

研究了单隐层前馈型小波神经网络模拟高度复杂的非线性映射。针对如何减小网络结构，提出了用小波元能量密度作为衡量其是滞是有效神经元的标准，并对网络构造，学习中的问题提出了观点。

Wavelet Transform and the Selection of Wavelet Neurons

In terms of wavelet analysis theory and Y. C. Pati's wavelet network, we propose a new notion of wavelet network for approximating arbitrary nonlinear functions. In this paper we introduce the conception of energy density which can be used as a standard to choose wavelet -neurons. The algorithms of network construction and learning are presented. We also compared our network with other wavelet-networks and BP networl, the experimental result is reported.It proved that the number of our network's wavelet-neurons is relatively small and the speed of learning is fast. We also bring forward our points on the problems produced in practice.