On the Multi-Stage Influence Maximization Problem

The influence maximization problem turns up in many online social networks (OSN) in which each participant can potentially influence the decisions made by others in the network. Relationships can be friendships, family relationships or even professional relationships. Such influences can be used to achieve some objective and the influence maximization problem attempts to make decisions so as to maximize the effect of these influences. Past work focused on a static problem whereby one tries to identify the participants who are the most influential. Recently, a multi-stage version of the problem was proposed in which outcomes of influence attempts are observed before additional participants are chosen. For example, in online advertising, one can provide an impression to a particular subject and if that subject clicks on the impression, then their friends are informed in the hope that this information will increase the chances that they also click on the impression and eventually purchase the product. This problem is computationally intensive; in this paper we investigate various optimization methods for finding its solution that yield close to optimal results while taking less computation time. These include greedy and particle swarm optimization algorithms.