Pseudo fuzzy clustering derived from Fisher criterions

This paper describes a new revised clustering algorithm in which each cluster center derived from the revised mean of a subclass in previous recursion. This modification factors make up with the mean of the cluster center in previous recursion multiplied with a coefficient polynomial. This computing center formula is derived from Fisher criteria. Experimental results show that the proposed clustering algorithm outperforms several other state of the art methods. It enjoys all advantage of K-means algorithm, and possesses faster running speed than kernel-based methods.