A Thorough Analysis of the Suppressed Fuzzy C-Means Algorithm

Suppressed fuzzy c-means (s-FCM) clustering was introduced in [Fan, J. L., Zhen, W. Z., Xie, W. X.: Suppressed fuzzy c-means clustering algorithm. Patt. Recogn. Lett. 24, 1607---1612 (2003)] with the intention of combining the higher speed of hard c-means (HCM) clustering with the better classification properties of fuzzy c-means (FCM) algorithm. They added an extra computation step into the FCM iteration, which created a competition among clusters: lower degrees of memberships were diminished according to a previously set suppression rate, while the largest fuzzy membership grew by swallowing all the suppressed parts of the small ones. Suppressing the FCM algorithm was found successful in the terms of accuracy and working time, but the authors failed to answer a series of important questions. In this paper we attempt to clarify the view upon the optimality and the competitive behavior of s-FCM via analytical computations and numerical analysis.