Abstract:The widely applied k-means algorithm produces clusterings that violate our expectations with respect to high/low similarity/density within/between clusters and is in conflict with Kleinberg’s axiomatic system for distance based clustering algorithms that formalizes those expectations. In particular, k-means violates the consistency axiom. We hypothesise that this clash is due to the unexplained expectation that the data themselves should have the property of being clusterable in order to expect the algorithm clustering them to fit a clustering axiomatic system. To demonstrate this, we introduce two new clusterability properties, i.e., variational k-separability and residual k-separability, and show that then Kleinberg’s consistency axiom holds for k-means operating in the Euclidean or non-Euclidean space. Furthermore, we propose extensions of the k-means algorithm that fit approximately Kleinberg’s richness axiom, which does not hold for k-means. In this way, we reconcile k-means with Kleinberg’s axiomatic framework in Euclidean and non-Euclidean settings. Besides contribution to the theory of axiomatic frameworks of clustering and to clusterability theory, the practical benefit is the possibility to construct datasets for testing purposes of algorithms optimizing the k-means cost function. This includes a method of construction of clusterable data with a global optimum known in advance.

A Proof of the Convergence Theorem of Maximum-Entropy Clustering Algorithm

ON THE COMPLETE CONVERGENCE OF U-STATISTICS

The Convergence of the Sums of Independent Random Variables under the Sub-Linear Expectations

Wide Gaps and Clustering Axioms

A remarkable example on clustering of extremes for regularly-varying stochastic processes

Wide gaps and Kleinberg’s clustering axioms for k -means

Bourgain's counterexample in the sequential convergence problem for the Schrödinger equation

Approximate maximum entropy principles via Goemans-Williamson with applications to provable variational methods

Monotonic Convergence in an Information-Theoretic Law of Small Numbers

On the Correct Convergence of the EM Algorithm for Gaussian Mixtures

Proof Of A Conjecture On Spearman'S P And Kendall'S Tau For Sample Minimum And Maximum

Research on Correct Convergence of the EM Algorithm for Gaussian Mixtures

Convergence analysis of a global optimization algorithm for centroid-based clustering

Can Evolutionary Clustering Have Theoretical Guarantees?

Convergence of a class of algorithms with the level-set subdifferential error bound

Entropy-Inspired Competitive Clustering Algorithms.

A counterexample on multiple convergence without commutativity

On entropic convergence of algorithms in terms of domain partitions

Kernel-based Maximal-Margin Clustring Algorithm

An Elementary Proof of the Inequality $χ\leq χ^*$ for Conditional Free Entropy

On the Convergence Rate of a Quasi-Newton Method for Inverse Eigenvalue Problems