On some tractable and hard instances for partial incentives and target set selection

TL;DR

This paper investigates the computational complexity of influence diffusion models in social networks, showing certain instances are hard while others are tractable, and introduces approximation algorithms for related problems.

Contribution

It proves hardness results for partial incentives and target set selection in specific graph classes, and provides efficient algorithms and approximations for related problems.

Findings

Finding optimal partial incentives is hard for chordal and planar graphs.

Tractable solutions exist for graphs with bounded treewidth and interval graphs with bounded thresholds.

An $O(\sqrt{n})$-approximation and a PTAS are developed for the dual problem of maximum degenerate set.

Abstract

A widely studied model for influence diffusion in social networks are {\it target sets}. For a graph $G$ and an integer-valued threshold function $\tau$ on its vertex set, a {\it target set} or {\it dynamic monopoly} is a set of vertices of $G$ such that iteratively adding to it vertices $u$ of $G$ that have at least $\tau(u)$ neighbors in it eventually yields the entire vertex set of $G$. This notion is limited to the binary choice of including a vertex in the target set or not, and Cordasco et al.~proposed {\it partial incentives} as a variant allowing for intermediate choices. We show that finding optimal partial incentives is hard for chordal graphs and planar graphs but tractable for graphs of bounded treewidth and for interval graphs with bounded thresholds. We also contribute some new results about target set seletion on planar graphs by showing the hardness of this problem, and by describing an efficient $O(\sqrt{n})$-approximation algorithm as well as a PTAS for the dual problem of finding a maximum degenerate set.