Algorithmic aspects of quasi-kernels

TL;DR

This paper investigates the computational complexity of finding small quasi-kernels in directed graphs, proving hardness results in general and polynomial-time algorithms for specific cases like orientations of trees.

Contribution

It establishes the difficulty of distinguishing graphs with small quasi-kernels from those without, and provides efficient algorithms for orientations of trees.

Findings

Computing small quasi-kernels is hard in general digraphs.

It is polynomial-time solvable for orientations of trees.

Counterexamples disprove the conjecture that all sink-free digraphs have small quasi-kernels.

Abstract

In a digraph, a quasi-kernel is a subset of vertices that is independent and such that every vertex can reach some vertex in that set via a directed path of length at most two. Whereas Chv\'atal and Lov\'asz proved in 1974 that every digraph has a quasi-kernel, very little is known so far about the complexity of finding small quasi-kernels. In 1976 Erd\H{o}s and Sz\'ekely conjectured that every sink-free digraph $D = (V, A)$ has a quasi-kernel of size at most $|V|/2$. Obviously, if $D$ has two disjoint quasi-kernels then it has a quasi-kernel of size at most $|V|/2$, and in 2001 Gutin, Koh, Tay and Yeo conjectured that every sink-free digraph has two disjoint quasi-kernels. Yet, they constructed in 2004 a counterexample, thereby disproving this stronger conjecture. We shall show that, not only sink-free digraphs occasionally fail to contain two disjoint quasi-kernels, but it is computationally hard to distinguish those that do from those that do not. We also prove that the problem of computing a small quasi-kernel is polynomial time solvable for orientations of trees but is computationally hard in most other cases (and in particular for restricted acyclic digraphs).