基于串联BP神经网络多函数拟合的研究设计

Research and design of multi-function fitting based on serial BP neural network

在研究神经网络对非线性函数拟合的过程中发现一个网络同时只能对一个函数进行拟合,为解决这一问题,通过研究BP算法神经网络提出了串联BP网络来实现多个非线性函数的拟合。通过Matlab自定义网络结构实现串联BP网络,在Matlab下利用此串联网络同时对两个非线性函数进行了拟合,训练样本加入噪声信号,在经过342次训练后拟合误差达到了0.000 978。实验测试结果显示本网络能够很好地对两个待拟合函数进行拟合,为同时进行多函数的拟合提供了新的方法。

It is found out in the study on nonlinear function fitting by neural network that one network can fit only one func-tion once. In order to solve this problem,a serial BP network is proposed in this paper to realize the multi-nonlinear function fit-ting on the basis of study on the BP algorithm neural network. The series BP network is realized by defining the network struc-ture by means of Matlab. This series network was adopted to fit the two nonlinear functions at the same time by means of Matlab. The noise signal was introduced into a training sample. After training of 342 times,fitting error reached 0.000 978. The testing results in experiments show that this network can fit two functions well,which provides a new method for multi-function simulta-neous fitting.