摘要
BP神经网络利用误差的反向传播调整神经网络的权值,BP神经网络的训练速度和训练误差很大程度上取决于学习速率和动量因子的设置。本文提出了一种改进的BP神经网络模型,学习速率和动量因子随误差实时调节,并进行了仿真。仿真结果表明,改进的BP神经网络比传统的BP神经网络收敛更快,误差更小。
BP neural network adjusts the weights by error back propagation, the training speed and error of the network are most depend on learning speed and momentum factor. In this paper, an improved BP neural network is proposed. The learning speed and momentum factor are adjusted online with the error change. The results of simulation show that this network converges faster.
出处
《自动化技术与应用》
2013年第9期7-9,共3页
Techniques of Automation and Applications
