deviation of fuzzy (or weighted) median as solution for cluster centers of the fuzzy-c-median algorithm

deviation of fuzzy (or weighted) median as solution for cluster centers of the fuzzy-c-median algorithm

Post by jo » Wed, 20 Aug 2003 05:05:06



In his article [1] the author mentioned, the fuzzy median is the
solution for the cluster centers of the fuzzy-c-median algorithm
("fuzzy-c-means" with l1-norm, calculated with absolute and not
squared distances between the objects). Unfortunately i can't follow
his argumentation in this point (is there any?).
Now i'm looking for the deviation or an explanation of this solution.
Thanks for any hint.

[1] Paul R. Kersten: Fuzzy Order Statistics and Their Application to
Fuzzy Clustering, IEEE Transactions on Fuzzy Systems, Vol. 7, NO. 6,
December 1999