Abstract

This paper views clusterings as elements of a lattice. Distances between clusterings are analyzed in their relationship to the lattice. From this vantage point, we first give an axiomatic characterization of some criteria for comparing clusterings, including the variation of information and the unadjusted Rand index. Then we study other distances between partitions w.r.t these axioms and prove an impossibility result: there is no “sensible” criterion for comparing clusterings that is simultaneously (1) aligned with the lattice of partitions, (2) convexely additive, and (3) bounded. 1.

Citations

6514	Elements of Information Theory - Cover, Thomas - 2004
991	Algorithmic Graph Theory and Perfect Graphs - Golumbic - 1980
675	RP: Enumerative Combinatorics I - Stanley - 1997
322	Arable P: Comparing partitions - Hubert - 1985
287	Objective criteria for the evaluation of clustering methods - Rand - 1971
156	Probability Theory - Rényi - 1970
105	Mathematical classification and clustering - Mirkin - 1996
83	A method for comparing two hierarchical clusterings - Fowlkes, Mallows - 1983
72	A stability based method for discovering structure in clustered data - Ben-Hur, Elisseeff, et al.
68	An impossibility theorem for clustering - Kleinberg
59	Comparing clusterings by the variation of information - Meila - 2003
24	S: Performance criteria for graph clustering and markov cluster experiments. Technical report INS-R0012, Centrum voor Wiskunde en Informatica 2000 - Dongen