论文部分内容阅读
提出了二进小波神经网络的结构及算法 ,并用于单组分和多组分示波计时电位信号的浓度计算。在二进小波神经网络中选用了Morlet母小波和修正的误差反传前向神经网络。探讨了二进小波神经网络中小波基个数、初始学习速率因子和动量因子等参数对网络预测结果的影响。结果表明 :二进小波神经网络对双组分和单组分示波计时电位信号中去极剂浓度的预测均有很好效果
The structure and algorithm of binary wavelet neural network are proposed and used to calculate the concentration of single-component and multi-component oscillographic chronopotentials. The Morlet mother wavelet and the modified error backpropagation neural network are selected in the binary wavelet neural network. The influence of parameters such as the number of wavelet bases, initial learning rate factor and momentum factor on the network prediction results in binary wavelet neural networks is discussed. The results show that the binary wavelet neural network is very effective in predicting the concentration of depolarizer in both component and one-component oscillographic chronopotentials