M-estimators and robust fuzzy clustering

Abstract

We compare several recent robust fuzzy clustering methods and show their equivalence in terms of their connection to techniques in robust statistics. In particular, we establish a connection between fuzzy set theory (as used in robust fuzzy clustering methods) and robust statistics, and point out the similarities between robust clustering methods and statistical methods such as the the M-estimator. We also give qualitative and quantitative equivalence of the robust fuzzy clustering techniques with M-estimator.