数值优化改进的BP神经网络逼近性能对比研究

Comparative study of approximation performance of numerical optimization improved BP neural networks

为了比较不同的数值优化改进的BP神经网络的逼近性能,本文在MATLAB 7.0环境下,建立了三类基于数值优化改进的BP算法,并以非线性函数逼近为例,对7种典型的数值优化改进算法进行网络训练和仿真实验,得出了在不同环境下,每种数值优化差法逼近的可行性。

We address numerical optimization improved BP neural network algorithms to compare their approximation capabilities.We establish three kinds of numerical optimization improved BP algorithms in MATLAB 7.0.Network training and simulation test are then performed for seven typical numerical optimization improved algorithms with non-linear function approximation as an example.We compare the approximation performance of different numerical optimization in defferent environments.