ON STOCHASTIC CONVERGENCE THEOREMS FOR THE FUZZY C-MEANS CLUSTERING PROCEDURE∗

The fuzzy c-means clustering algorithm has been well studied for equal weight distributions on a finite set. Suppose that this situation is generalized to an arbitrary probability distribution on a finite dimensional Euclidean space, assuming that the second moment of the distribution is finite. Now choose ever larger finite random samples from this distribution and compute the standard optimal membership functions for a fuzzy partition into c clusters. Then the convergence of the cluster center points is established in the Hausdorff sense with probability one, provided that there is a unique optimal center point set. These optimal center points are the fixed point of a simple operator, and there is a corresponding iterative algorithm that generalizes the usual procedure.

[1]  Michael J. Sabin,et al.  Convergence and Consistency of Fuzzy c-means/ISODATA Algorithms , 1987, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[2]  James C. Bezdek,et al.  A Convergence Theorem for the Fuzzy ISODATA Clustering Algorithms , 1980, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[3]  W. T. Tucker,et al.  Convergence theory for fuzzy c-means: Counterexamples and repairs , 1987, IEEE Transactions on Systems, Man, and Cybernetics.

[4]  James C. Bezdek,et al.  Pattern Recognition with Fuzzy Objective Function Algorithms , 1981, Advanced Applications in Pattern Recognition.

[5]  J. Bezdek,et al.  FCM: The fuzzy c-means clustering algorithm , 1984 .

[6]  D. Pollard.  Strong Consistency of $K$-Means Clustering , 1981 .

[7]  J. C. Dunn,et al.  A Fuzzy Relative of the ISODATA Process and Its Use in Detecting Compact Well-Separated Clusters , 1973 .

[8]  T. Apostol.  Mathematical Analysis , 1957 .