关键词: 数值优化, BP神经网络, 逼近性能, 对比研究

Abstract:

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.

Key words: numerical optimization, BP neural networks, approximation performance, comparative study