用U.L.N.神经网络实现的滤波器

FILTERS BUILT BY U.L.N. NEURAL NETWORK

本文的工作是探讨了用三层 U .L .N .神经网络实现低通 ,高通和带通滤波器的方法 .得到了以上三种滤波器在阻带 ,通带和过渡带与逼近函数之间的误差均低于 10 - 1 6的结果 .以实例证明了三层的 U.L.N.神经网络不仅可以实现任意的函数关系 ,而且可以实现具有特定频率特性的滤波器功能 .

This paper have researched the method of designing lowpass, highpass and bandpass filters by U.L.N.neural network. The result is the errores are less than 10. It demonstrates that U.L.N. network can not only realize normal functions, but also realize filters.