Title:Categorization Axioms for Classification and Clustering

Abstract:Categorization involves classification and clustering. In this paper, we first propose a united representation of categorization results and then present an axiomatic framework for categorization by assuming that a proper categorization result should satisfy categorization equivalency axiom, sample separation axiom and category separation axiom. The proposed axiomatic framework is not only consistent with categorization theories in cognitive science, but also classifies categorization results and partitions. The proposed separation axioms can lead to three principles of designing categorization methods. And classification and clustering are theoretically investigated based on categorization axioms.