色谱指纹图谱识别的人工神经网络方法对比研究

Comparison of Three Artificial Neural Network Methods Applied in Traditional Chinese Medicine Chromatographic Fingerprint

将反传人工神经网络、径向基人工神经网络和自组织竞争人工神经网络方法用于清热解毒口服液色谱指纹图谱的解析。建立了清热解毒口服液色谱指纹图谱三种人工神经网络模型,采用留一法评价三种网络的学习能力,结果表明,与反传人工神经网络和径向基人工神经网络相比,自组织竞争人工神经网络有更强的自学习能力和泛化能力。

In the paper, three typical artificial neural networks, including back propagation network, Radial basis network and Self-organized competition network, were employed to analyze the high performance liquid chromatographic fingerprint data of Qing-Re-Jie-Du Oral Liquid. Chromatographic fingerprint of three artificial neural network model were established and leave-one-out method was applied to evaluate their ability of studing and generalizing. It was found that Self-organized competition network not only has strong ability of self-studing, but also stronger ability of generalizing than other two kinds of networks.