摘要
为了优化BP神经网络,提出了一种优化BP神经网络的流程。首先,判断各影响因素之间的自相关性,如果各影响因素满足自相关评价指标,则可以使用BP神经网络进行回归训练;其次,改变BP神经网络的隐藏节点数、学习效率、训练误差和训练次数等影响因素;最后,加入遗传算法或者粒子群算法与BP神经网络组成混合算法,以提高BP神经网络的训练精度。
In order to optimize the BP neural network,this paper proposes a process to optimize the BP neural network.First of all,judge the auto-correlation between each influencing factor.If each influencing factor meets the auto-correlation evaluation index,the BP neural network can be used for regression training.Secondly,change the number of hidden nodes,learning efficiency,training error,training times and other influencing factors of BP neural network.Add a genetic algorithm or particle swarm algorithm and BP neural network to form a hybrid algorithm to improve the training accuracy of BP neural network.
作者
何大四
金璐琪
张祖铭
赵强强
HE Dasi;JIN Luqi;ZHANG Zuming;ZHAO Qiangqiang(School of Energy&Environment,Zhongyuan University of Technology,Zhengzhou 451191,China)
出处
《机械工程与自动化》
2025年第1期224-226,共3页
Mechanical Engineering & Automation
关键词
BP神经网络
隐藏节点
混合算法
回归预测
自相关性
BP algorithm
hidden node
hybrid algorithm
regression prediction
auto-correlation
