人工神经网络用于中药材雷公藤和昆明山海棠的分类识别研究

CLASSIFICATION AND RECOGNITION OF TRIPTERYGIUM WILFORDII AND T.HYPOGLAUCUM BY ARTIF101AL NEUKAL NETWORKS

应用误差反向传播学习算法（ＢＰ算法）对中药材雷公藤和昆明山海棠浸出物的红外光谱进行分类识别。网络为３层结构，输入节点为９个，隐层节点为２１个，输出节点为１。分类结果与ＸＩＭ－ＣＡ法基本一致。此外，本文还考察了网络参数间的相互关系。

A BASIC program for the simulation of artificial neural networks was implementedon a 386SX. For classification and recognition of infrared(IR) spectra of extracts by means of aneural network, a back propagation model with one hidden layer and a sigmoid transfer function hasbeen proved to be available. The nine features selected by the Shannon information content were usedas input elements. The hidden layer contains 21 nodes and the output layer contains one node.