一种用于非线性函数逼近的小波神经网络

A kind of wavelet neural networks used in approaching non-linear functions

提出一种用于非线性函数逼近的小波神经网络 ,给出了网络的参数训练方法。从信息熵的概念出发 ,改进了网络参数训练的目标函数 ,并利用引入动量项的最速下降法训练网络权值、尺度因子和平移因子。仿真实验表明 ,该小波神经网络用于非线性函数逼近时优于同等规模的BP网络 ,且其训练方法亦具有收敛速度快、逼近精度高等优点。

A kind of wavelet neural networks used in approaching non-linear functions is proposed,and the method of parameter study of the network is given.By the conception of information entropy,we proposed a new improved objective function,the weights of the network,scale factor and displacement factor are studied by the steepest descent method.The results of simulation indicate that the wavelet neural networks algorithm is better than the BP neural networks when used in approaching non-linear functions.The studied method is fast in its convergence speed and has a good approaching precision.