Complexity of Local Search for Euclidean Clustering Problems
Computational Complexity
2025-05-12 v3 Data Structures and Algorithms
Abstract
We show that the simplest local search heuristics for two natural Euclidean clustering problems are PLS-complete. First, we show that the Hartigan--Wong method for -Means clustering is PLS-complete, even when . Second, we show the same result for the Flip heuristic for Max Cut, even when the edge weights are given by the (squared) Euclidean distances between the points in some set ; a problem which is equivalent to Min Sum 2-Clustering.
Cite
@article{arxiv.2312.14916,
title = {Complexity of Local Search for Euclidean Clustering Problems},
author = {Bodo Manthey and Nils Morawietz and Jesse van Rhijn and Frank Sommer},
journal= {arXiv preprint arXiv:2312.14916},
year = {2025}
}
Comments
29 pages, 4 figures