摘要
以结构为 3- 5 - 2的网络为例 ,对BP神经网络的稳定性、权值的收敛趋势、样本识别率和网络总误差对权值的影响等做了研究 ,结果表明系统总误差的减小会促使网络权值将向最优方向调整 ,但这种调整有一定极限 .
In this paper,research on the stability,the convergence tendency of the power value,the sample identification rate of the BP Neural Network and the influence to the power value was made by the Network's general error with a example whose structure of the BP Neural Network is 3-5-2.We draw some meaningfull conclusions and correlated data here.
出处
《滨州师专学报》
2000年第4期34-37,57,共5页
Journal of Binzhou Teachers College
关键词
BP神经网络
权值
样本识别率
网络总误差
BP neural network,power value,sample identification rate ,the network's general errx
