A Convergence Theorem for Improved Kernel Based Fuzzy C-Means Clustering Algorithm
Fuheng Qu, Yating Hu, Yong Yang, Shuangzi Sun · 2011
In 2008, we proposed a clustering algorithm called improved kernel based fuzzy c-means clustering algorithm (IKFCM) to improve the performance of the original fuzzy c-means clustering algorithm. In this paper, we analyze the convergence of the IKFCM by means of Zangwill's convergence theorem. The result shows that arbitrary sequences generated by IKFCM always terminates at a local minimum or saddle point, or at worst, al-ways contains a subsequence which converges to a local minimum or saddle point of the IKFCM clustering model.