Weighting Exponent Selection of Fuzzy C-Means via Jacobian Matrix

摘要

FCM is a popular clustering algorithm and applied in various areas. However, there are still some problems to be solved including the selection of weighting exponent m and convergence analysis. In this paper, we present an efficient method to identify the proper range of m and convergence rate by a new Jacobian matrix of FCM. A series of experimental results on both synthetical data and real-world data validate the proposed theoretical results.