BP神经网络算法的改进及在Matlab中的实现

DESIGN AND REALIZE THE SOFTWARE OF MACHINING TECHNIQUES MANUAL

分析了BP神经网络算法收敛速度慢的原因,研究了一种提高其收敛速度的方法,即将动量法和批处理方法结合起来对BP算法进行改进,提高学习速度,并给出了算法思想,用Matlab神经网络工具箱对神经网络进行了初始化和训练,仿真测试结果证明该方法是可行,具有实用性。

This paper analysis the reason why the algorithm of BP is slowly convergent, then introduced a method of improving arithmetic on BP neural networks which absorbed adding momentum ways and batch ways to increase its study speed and astringency. Then applying Matlab to initialize and train BP neural networks. The result proves the method is feasible.