基于监督模糊ART的中药材模式识别和质量评价

为实现基于高效液相色谱的中药材识别和质量评价，提出一种有监督ART神经网络。有监督模糊ART神经网络与模糊ART相比主要有两处改进，一是改为有监督，二是为每个聚类（子类）中心单独设置了警戒门限，从而使网络成为具有子类自动聚类能力的有监督模糊ART神经网络。该网络与模糊ART具有一样的性能和优点，比如添加新类不须重新学习等。因为中药材会由于 产地、采摘时间和处理方法的不同会有较大的区别，所以子类的自动聚类能力对中药材模式识别尤其有意义。经大量中药材样本和实际测试表明，网络抗平移、形变和适应新产地药材的能力都很强，达到了预期效果。

Traditional Chinese Medicine Pattern Recognition and Quality Evaluation Based on a Supervised Fuzzy ART

A supervised fuzzy ART is presented for recognizing traditional Chinese medicines and evaluating its quality based on high performance liquid chromatogram(HPLC). It has two improvements to a fuzzy ART--is turn first to a supervised fuzzy ART then is set a single vigilance parameter for every clustering-center (sub-class), therefore a fuzzy ART becomes a supervised fuzzy ART with the automatic clustering ability to sub-class. The supervised fuzzy ART have the performances and advantages as the same as a fuzzy ART, for example, the network doesn't learn again while a new class must be added to it. The automatic clustering ability to sub-class has great meaning to the pattern recognition of traditional Chinese medicine because the traditional Chinese medicine will be very different under the condition of different place come from, different time picked up and different method dealt with. It is manifested after abundant of traditional Chinese medicine samples have been used to test that the network's abilities, such as the resistance to parallel removal, deformation and the adaptability to the medicines coming from new places, are very strong. So, it indicts the arrival to the prospective.