Combinatorial Optimization Games Arise in Social Networks Team Members :

Rui Zhang,M. Sahin

2014-01-01

Abstract:Recently, the dynamic processes for the diffusion of influence has attracted significant interest from algorithmic researchers. Over the past ten years, the following viral marketing problem has attracted a significant amount of interest. Assume we want to promote a new product over a given social network and wish this product will be adopted by most people in this network. We can initialize the diffusion process by “targeting” some influential people. Then, a cascade will be caused by these initial adopters and other people start to adopt this product due to the influence they receive from earlier adopters. But how should we select these influential people who are targeted initially? Domingos and Richardson [2001] studied the problem in a probabilistic setting, and provided heuristic solutions. Kempe et al. [2003] were the first ones to model this problem as an optimization problem by using the threshold model proposed by Granovetter [1973], showed it is NP-hard to find the optimal initial set, and developed approximation algorithms for the problem. Subsequently, several different variants of this problem have been studied. Since then, several variants of this problem are studied. Among them, Chen [2009] presented the Target Set Selection (TSS) problem. Built on the TSS problem, Raghavan and Zhang [2013] proposed the Weighted Target Set Selection (WTSS) problem and Gunnec et al. [2013] studied the Least Cost Influence Problem (LCIP). The WTSS problem is defined on a connected undirected graph G = (V,E). The node set V can be interpreted as a set of people in a social network then the edge set E becomes the connections between people on the social network. For each i ∈ V , there is a threshold, denoted by gi, which is between 1 and α(i), where α(i) denotes the degree of node i. Let gi and bi be the threshold value and the weight for node i ∈ V respectively. The weight models the fact that different nodes require differing levels of effort to become initial adopters (in practice, it is reasonable to assume different persons would not always require the same amount of effort to be convinced). Initially we assume all nodes are inactive. We select a subset