摘要
本文将变尺度法引入到神经网络的学习中。在修改权值时 ,引入串行修正权值的思想。基于以上两点提出的算法具有收敛速度快。
This paper introduces the variable-schedule method into the training of back-propagation neural network. In the process of modifing weights, the paper introduces the way of series. Based the improvement, the algorithm has the fast convergence speed and the high accuracy.
出处
《微机发展》
2001年第4期50-51,53,共3页
Microcomputer Development
