Parameterized algorithms for locating-dominating sets

TL;DR

This paper investigates the parameterized complexity of the locating-dominating set problem, establishing tight bounds on algorithmic efficiency and kernelization for various structural graph parameters.

Contribution

It provides new complexity bounds, kernelization results, and open problems for the locating-dominating set problem under different graph parameters.

Findings

No subexponential algorithm under ETH for solution size parameterization.

Exponential kernel for distance to cluster parameterization.

Linear kernel for max leaf number parameterization.

Abstract

A locating-dominating set $D$ of a graph $G$ is a dominating set of $G$ where each vertex not in $D$ has a unique neighborhood in $D$, and the Locating-Dominating Set problem asks if $G$ contains such a dominating set of bounded size. This problem is known to be $\mathsf{NP-hard}$ even on restricted graph classes, such as interval graphs, split graphs, and planar bipartite subcubic graphs. On the other hand, it is known to be solvable in polynomial time for some graph classes, such as trees and, more generally, graphs of bounded cliquewidth. While these results have numerous implications on the parameterized complexity of the problem, little is known in terms of kernelization under structural parameterizations. In this work, we begin filling this gap in the literature. Our first result shows that Locating-Dominating Set, when parameterized by the solution size $d$, admits no $2^{o(d \log d)}$ time algorithm unless the Exponential Time Hypothesis fails; as a corollary, we also show that no $n^{o(d)}$ time algorithm exists under ETH, implying that the naive $\mathsf{XP}$ algorithm is essentially optimal. We present an exponential kernel for the distance to cluster parameterization and show that, unless $\mathsf{NP-hard} \subseteq \mathsf{NP-hard}/$\mathsf{poly}$, no polynomial kernel exists for Locating-Dominating Set when parameterized by vertex cover nor when parameterized by distance to clique. We then turn our attention to parameters not bounded by neither of the previous two, and exhibit a linear kernel when parameterizing by the max leaf number; in this context, we leave the parameterization by feedback edge set as the primary open problem in our study.