提高BP网络学习速度的自适应算法

针对BP网络存在收敛速度慢及目标函数容易陷入局部极小值的缺点,本文首先对前人所做的改善BP网络学习速度的方法进行了探讨,其次分析了学习率对BP网络学习速度的影响,最后提出了一种提高BP网络学习速度的新方法,即自适应调整学习率和动量因子.计算结果表明,该方法大大地提高了收敛速度,而且算法简单、易行.

Adaptive Algorithm to Improve Learning Speed of BP Network

s:In order to solve the problem that BP network has the disadvantages of slow convergence speed and target function getting into local infinitesimal value, this paper firstly discusses some ways to improve learning speed of BP network what have been done by foreman. Secondly, it analyzes how the learning rate affects the learning speed of the BP network. Lastly, it presents a mew method improving learning speed of the BP network. The method is to adjust the learning rate and the momentum gene self-adaptively. The results of calculation showed that this method not only improves the convergence speed greatly, but also be simple and feasible.