Representable hierarchical clustering methods for asymmetric networks

Abstract

This paper introduces the generative model of representability for hierarchical clustering methods in asymmetric networks, i.e., the possibility to describe a method through its action on a collection of networks called representers. We characterize the necessary and sufficient structural conditions needed on these representers in order to generate a method which is scale preserving and admissible with respect to two known axioms and, based on this result, we construct the family of cyclic clustering methods.