Abstract: The main weak point of Back Propagation(BP) algorithm is that the optimal procedure is easily trapped into local minimum value and the speed of convergence is very slow. To solve the problem, this paper makes use of ergodicity property of chaos, starts its improvement from the learning rate. The improved algorithm undergoes a simulated operation with Matlab. The outcome shows that the algorithm improves the speed of network study and the accuracy of convergence, and saves the network from the problem of local minima.

Key words: Back Propagation(BP) neural network, chaos, learning rate, ergodicity property

摘要： 针对BP算法在训练过程中容易陷入局部极小值，导致收敛速率慢的问题，探讨一种利用混沌的遍历特性改进学习效率的算法，用Matlab软件对改进算法进行仿真。实验结果表明，该算法能够提高神经网络的学习效率和收敛精度，较好地避免网络陷入局部极小点。

关键词: BP神经网络, 混沌, 学习率, 遍历性

CLC Number: 

• TP311