摘要
该文讨论了BP网络学习过程中的假饱和现象 ,并给出了一种改进的算法 ,有效地解决了假饱和的问题。仿真结果表明 ,该方法不但可以提高网络学习的快速性 ,而且具有一定的避免权值落入局部极小点的能力 ,从而提高了网络的收敛精度 ,同时 。
This paper discusses fault saturation in the studing process of BP network,and presents an improved algorithm which can avoid saturation. The simulation results prove that the algorithm can not only complete the study process quickly, but also avoid local minimum in studying process of neural network. So the algorithm can also improve the precision of the neural network and the generalization capability.
出处
《计算机仿真》
CSCD
2004年第3期82-84,共3页
Computer Simulation
