English

A Fast Approximation Scheme for Low-Dimensional $k$-Means

Data Structures and Algorithms 2017-08-30 v2 Computational Geometry

Abstract

We consider the popular kk-means problem in dd-dimensional Euclidean space. Recently Friggstad, Rezapour, Salavatipour [FOCS'16] and Cohen-Addad, Klein, Mathieu [FOCS'16] showed that the standard local search algorithm yields a (1+ϵ)(1+\epsilon)-approximation in time (nk)1/ϵO(d)(n \cdot k)^{1/\epsilon^{O(d)}}, giving the first polynomial-time approximation scheme for the problem in low-dimensional Euclidean space. While local search achieves optimal approximation guarantees, it is not competitive with the state-of-the-art heuristics such as the famous kk-means++ and D2D^2-sampling algorithms. In this paper, we aim at bridging the gap between theory and practice by giving a (1+ϵ)(1+\epsilon)-approximation algorithm for low-dimensional kk-means running in time nk(logn)(dϵ1)O(d)n \cdot k \cdot (\log n)^{(d\epsilon^{-1})^{O(d)}}, and so matching the running time of the kk-means++ and D2D^2-sampling heuristics up to polylogarithmic factors. We speed-up the local search approach by making a non-standard use of randomized dissections that allows to find the best local move efficiently using a quite simple dynamic program. We hope that our techniques could help design better local search heuristics for geometric problems. We note that the doubly exponential dependency on dd is necessary as kk-means is APX-hard in dimension d=ω(logn)d = \omega(\log n).

Keywords

Cite

@article{arxiv.1708.07381,
  title  = {A Fast Approximation Scheme for Low-Dimensional $k$-Means},
  author = {Vincent Cohen-Addad},
  journal= {arXiv preprint arXiv:1708.07381},
  year   = {2017}
}
R2 v1 2026-06-22T21:22:38.374Z