English

Deterministic Clustering in High Dimensional Spaces: Sketches and Approximation

Data Structures and Algorithms 2023-10-09 v1

Abstract

In all state-of-the-art sketching and coreset techniques for clustering, as well as in the best known fixed-parameter tractable approximation algorithms, randomness plays a key role. For the classic kk-median and kk-means problems, there are no known deterministic dimensionality reduction procedure or coreset construction that avoid an exponential dependency on the input dimension dd, the precision parameter ε1\varepsilon^{-1} or kk. Furthermore, there is no coreset construction that succeeds with probability 11/n1-1/n and whose size does not depend on the number of input points, nn. This has led researchers in the area to ask what is the power of randomness for clustering sketches [Feldman, WIREs Data Mining Knowl. Discov'20]. Similarly, the best approximation ratio achievable deterministically without a complexity exponential in the dimension are Ω(1)\Omega(1) for both kk-median and kk-means, even when allowing a complexity FPT in the number of clusters kk. This stands in sharp contrast with the (1+ε)(1+\varepsilon)-approximation achievable in that case, when allowing randomization. In this paper, we provide deterministic sketches constructions for clustering, whose size bounds are close to the best-known randomized ones. We also construct a deterministic algorithm for computing (1+ε)(1+\varepsilon)-approximation to kk-median and kk-means in high dimensional Euclidean spaces in time 2k2/εO(1)poly(nd)2^{k^2/\varepsilon^{O(1)}} poly(nd), close to the best randomized complexity. Furthermore, our new insights on sketches also yield a randomized coreset construction that uses uniform sampling, that immediately improves over the recent results of [Braverman et al. FOCS '22] by a factor kk.

Keywords

Cite

@article{arxiv.2310.04076,
  title  = {Deterministic Clustering in High Dimensional Spaces: Sketches and Approximation},
  author = {Vincent Cohen-Addad and David Saulpic and Chris Schwiegelshohn},
  journal= {arXiv preprint arXiv:2310.04076},
  year   = {2023}
}

Comments

FOCS 2023. Abstract reduced for arxiv requirements

R2 v1 2026-06-28T12:42:21.265Z