摘要
为提高BP神经网络的映射能力,通过反馈环节中权值阈值和输出误差来给调整公式中常数项动量因子和学习速率赋新值,提出改进BP神经网络算法。改进后的BP神经网络从理论上调整精度更高,训练速率更快,二者从原来的依靠经验赋值变成变量,训练适应性更强。通过仿真实验,相比传统BP神经网络,改进后的BP神经网络在训练速率上更快,预测精度明显提高。因此得出结论,改进后的BP神经网络算法在部分电子部件性能预测中具有更好的适用性。
In order to improve the mapping ability of BP neural network,the new values of the constant term momentum factor and the learning rate in the adjustment formula are given by the weight threshold and output error in the feedback link,and the improved BP neural network algorithm is proposed.The improved BP neural network has higher adjustment accuracy in theory and faster training speed.The two are changed from the original dependent on experience assignment into variables,and the training adaptability is stronger.Simulation results show that the improved BP neural network has faster training rate and higher prediction accuracy,compared with the traditional BP neural network.Therefore,it is concluded that the improved BP neural network algorithm has better applicability in the performance prediction of some electronic components.
作者
岳炯
吕卫民
苏宁远
柴志君
YUE Jiong;LV Weimin;SU Ningyuan;CHAI Zhijun(Naval Aviation University,Yantai 264001)
出处
《舰船电子工程》
2020年第3期111-115,123,共6页
Ship Electronic Engineering
关键词
BP神经网络
算法改进
动量因子
学习速率
性能预测
BP neural network
algorithm improvement
momentum factor
learning rate
performance prediction
