Dynamic monopolies in directed graphs: the spread of unilateral influence in social networks

TL;DR

This paper studies the spread of influence in directed social networks through dynamic monopolies, providing bounds, complexity results, and an efficient algorithm for the strict majority case.

Contribution

It establishes new upper bounds and polynomial-time algorithms for the minimum size of dynamic monopolies in directed graphs with strict majority thresholds.

Findings

Any directed graph with positive minimum in-degree has a strict majority dynamic monopoly of size at most n/2.

The n/2 bound can be achieved by a polynomial-time algorithm.

The paper improves previous bounds significantly.

Abstract

Let $G$ be a directed graph such that the in-degree of any vertex $G$ is at least one. Let also ${\mathcal{\tau}}: V(G)\rightarrow \Bbb{N}$ be an assignment of thresholds to the vertices of $G$. A subset $M$ of vertices of $G$ is called a dynamic monopoly for $(G,\tau)$ if the vertex set of $G$ can be partitioned into $D_0\cup... \cup D_t$ such that $D_0=M$ and for any $i\geq 1$ and any $v\in D_i$, the number of edges from $D_0\cup... \cup D_{i-1}$ to $v$ is at least $\tau(v)$. One of the most applicable and widely studied threshold assignments in directed graphs is strict majority threshold assignment in which for any vertex $v$, $\tau(v)=\lceil (deg^{in}(v)+1)/2 \rceil$, where $deg^{in}(v)$ stands for the in-degree of $v$. By a strict majority dynamic monopoly of a graph $G$ we mean any dynamic monopoly of $G$ with strict majority threshold assignment for the vertices of $G$. In this paper we first discuss some basic upper and lower bounds for the size of dynamic monopolies with general threshold assignments and then obtain some hardness complexity results concerning the smallest size of dynamic monopolies in directed graphs. Next we show that any directed graph on $n$ vertices and with positive minimum in-degree admits a strict majority dynamic monopoly with $n/2$ vertices. We show that this bound is achieved by a polynomial time algorithm. This upper bound improves greatly the best known result. The final note of the paper deals with the possibility of the improvement of the latter $n/2$ bound.