On the minimum of the mean-squared error in 2-means clustering

DOI10.2140/INVOLVE.2019.12.301zbMATH Open1406.62067arXiv1710.09693OpenAlexW2767194928WikidataQ129092160 ScholiaQ129092160MaRDI QIDQ1624878FDOQ1624878

Authors: Bernhard G. Bodmann, Craig J. George 

Publication date: 16 November 2018

Published in: Involve (Search for Journal in Brave)

Abstract: We study the minimum mean-squared error for 2-means clustering when the outcomes of the vector-valued random variable to be clustered are on two touching spheres of unit radius in $n$-dimensional Euclidean space and the underlying probability distribution is the normalized surface measure. For simplicity, we only consider the asymptotics of large sample sizes and replace empirical samples by the probability measure. The concrete question addressed here is whether a minimizer for the mean-squared error identifies the two individual spheres as clusters. Indeed, in dimensions $nge3$, the minimum of the mean-squared error is achieved by a partition that separates the two spheres and has unit distance between the points in each cluster and the respective mean. In dimension $n=2$, however, the minimizer fails to identify the individual spheres; an optimal partition is obtained by a separating hyperplane that does not contain the point at which the spheres touch.

Full work available at URL: https://arxiv.org/abs/1710.09693

Recommendations

• scientific article; zbMATH DE number 6902590
• Classification performance resulting from a 2-means
• Robust \(k\)-means clustering for distributions with two moments
• scientific article
• Good (K-means) clusterings are unique (up to small perturbations)

zbMATH Keywords

\(k\)-means clusteringperformance guaranteesmean-squared error

Mathematics Subject Classification ID

Nonparametric estimation (62G05) Classification and discrimination; cluster analysis (statistical aspects) (62H30)

Cites Work

• Least squares quantization in PCM
• Foundations of quantization for probability distributions
• A central limit theorem for k-means clustering
• Hypercontractivity for the heat semigroup for ultraspherical polynomials and on the n-sphere
• Multidimensional asymptotic quantization theory with<tex>r</tex>th power distortion measures
• \(k\)-means requires exponentially many iterations even in the plane
• Centroidal Voronoi Tessellations: Applications and Algorithms
• K-Means-Type Algorithms: A Generalized Convergence Theorem and Characterization of Local Optimality
• Exponential rate of convergence for Lloyd's method I
• Approximating K‐means‐type Clustering via Semidefinite Programming
• Probably certifiably correct \(k\)-means clustering

Cited In (2)

• The ratio-cut polytope and K-means clustering
• Minimum adjusted Rand index for two clusterings of a given size

This page was built for publication: On the minimum of the mean-squared error in 2-means clustering

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q1624878)