Fast Semi-Supervised Fuzzy Clustering: Approach and Application
Jiaxin Cai, Feng Yang, Guo-can Feng · 2009
This paper proposes a novel fast-semi-supervised-FCM algorithm (fsFCM) to fundamentally overcome the critical disadvantages of Pedrycz's semi-supervised-FCM(sFCM) ,i.e., degeneracy to classical FCM and slow convergence, particularly when applied in actual data set. Experimental results demonstrate that fsFCM can outperform sFCM in accuracy, speed and robustness for clustering. Moreover, it shows that fsFCM avoids the problems of slow convergence and degeneracy to FCM when applied to actual data clustering, and also presents its effectiveness for the application in medical images segmentation.