ART2网络结构与算法的改进

MART2:Modification in Structure and Algorithms

ART2网络中的模式匹配过程,以及相似度警戒测试过程均以模式的相似性量度值为基础。传统ART2网络的相似量度是一种关于模式相位信息的量度,在需要考虑模式幅度信息及处理集群分布样本时,效果很不理想。文章针对此不足,提出以欧氏距离为相似测度的新型网络:MART2。输入模式的幅度信息被提取出来,并送到相应的中间模式和警戒测试部分。新网络中引入三个辅助函数共同计算输入模式与存贮模式的相似度,使得在进行模式匹配和警戒测试时,幅度信息没有丢失。实验证明,MART2在处理集群分布样本时,性能优于传统ART2。MART2是对ART2网络的一种补充。

In an ART2Net,processes of patter n-matching and vigilance-testing must be based on the value of similari-ty of two patterns.The measurement of similarity in classical ART2is something abo ut the phases of pattern vectors.It doesn't work well when must take the magni tude of a pattern into consider,or when process a great of samples which distri bute as clustering groups.Based on this,the authors propose a new modified ART 2version:MART2.In the new one,the magnitude of an input patter is extracte d before the pattern being sent to STM-F1,and keep working during calcu-lat ing mediate input pattern and vigilance testing.We have introduced three spec ial functions together to help measure the similarity between an input pattern and the pattern stored in weight vector.The magnitude of the input pattern do esn't be lost through the whole matching and testing processes.Experiment in t his paper showed that the new MART2performed better than the classical one when grouping clustering data.MART2is a useful supplement and extension of class ical ART2family.It will promote further development in application of ART2Ne t.