一种用于函数学习的小波神经网络

在非线性系统辨识中,系统输入往往是多变量的.小波处理此类问题则比较复杂.结合神经网络形式和小波特点建立一种新型的网络,可简单有效地解决网络多输入问题.同时给出此网络可以逼近任意连续函数的数学证明.并通过实例验证了此方法的正确性.

Wavelet Neural Network for Function Learning

In identification of nonlinear system,thenetwork inputs usually are multivariab les.The approach using wavelets alone to resolvethe problem is complex.Therefore,the paper establishes a new network method by combiningneural network and wav elet characteristics to handle the multivariable problem simply andeffectively. This paper also gives the mathematical verification that the network canapproxi mate any kind of functions.The simulation results of an examkple indicate thevalidity of the method.