摘要
把模糊集合论中隶属度的概念引入SOM网络的学习过程中,用“走着瞧”的分类策略代替“一步定终身”的作法,较好地克服了SOM网络中模式类别分界线难以确定以及网络权值局部振荡而引起的学习时间延长的缺陷.该算法中隶属度由文献[1]中模糊ISODATA算法确定,本文将SOM与模糊ISODATA的这种结合称为模糊SOM网络,即FSOM.将其应用于实际的油气预测中,取得了良好的应用效果.
In this article, subordinate degree of Fuzzy Set Theory is introduced into the studying process in network SOM. Through the substitution of the strategy 'walk and see' for 'winner takes all', we conquer some drawbacks in Network SOM, such as the difficulty of determining the dividing line between pattern category and the delay of study time resulting from local vibration of network weigh. The subordinate degree in this algorithm is defined by Fuzzy ISODATA in document So we call it Fuzzy SOM, namely FSOM. We get a satisfactory result through applying it into practical oil-gas prediction.
出处
《河北工业大学学报》
CAS
2003年第5期100-103,共4页
Journal of Hebei University of Technology
