Generalized noise clustering as a robust fuzzy c-M-estimators model

Dave's noise clustering (NC) algorithm has been generalized in an earlier work, where the noise distance delta is allowed to take different values for different feature vectors. Based on that, it was shown that the membership generated by the NC algorithm is a product of two terms, one is the original fuzzy c-means (FCM) membership responsible for data partitioning, and the other is a generalized possibilistic membership that achieves a mode seeking effect and imparts robustness. In this paper, it is shown that a variety of robust M-estimators can be incorporated into the generalized NC algorithm, for example Huber, Hampel, Cauchy, Tukey biweight, and Andrew's sine. The generalized NC algorithm is also compared with the recently introduced mixed c-means clustering and a noise resistant FCM technique.