摘要
模糊协方差聚类算法实质是一种局部寻优搜索方法,其收敛结果易陷入局部极小.本文结合分级聚类的思想,提出了一种改进算法.实验结果表明改进算法得到最优解的比例提高了.
Fuzzy covariance clustering is an extension to fuzzy C-means algorithm that is essentially a partially optimization searching usually leading to local minimum results. In this paper, an approach of merging operations is presented for redundant initialization, so as to inherit the advantages of non-local minima of hierarchical clustering techniques, and overcome its shortage of being static. Two proposed merging criteria based on similarity information laid in fuzzy partition matrix are tested on simulated data set with Gaussian distribution and on line-circle mixture set, which show the significant improvement in converging to the global optimal solutions.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
2002年第3期322-325,共4页
Pattern Recognition and Artificial Intelligence
基金
安徽省教育厅科学基金(2000J1023)
