|CANTONI RICCARDO||Cycle: XXXV |
Section: Systems and Control
Tutor: PIRODDI LUIGI
Advisor: AMALDI EDOARDO Major Research topic
:Optimization models and methods for contrasting the spread of rumor in social networksAbstract:
Optimization in Social Networks is a hot topic that has been gaining increasing attention in the recent years. Most of the works in the literature are related to the Influence Maximization Problem (IMP), i.e. the problem of maximizing the spread of single information in social networks generated by a set of influencers. In this problem, the network is represented with a graph, and the spread of information is described by a propagation model. Heuristic algorithms, with
or without a provable approximation guarantee, and exact methods have been proposed to solve different versions of this problem, for example with different objective functions like maximizing the influence given a budget or minimizing the cost to achieve a given level of diffusion. Recently, the IMP has been extended to a multi-agent setting via the so-called Competitive Influence Maximization Problems (CIMP) with applications, among others, in viral marketing. In a CIMP each agent competes against the others generating an information cascade to persuade most of the network nodes to adopt its products. An emerging topic in this area is that of contrasting the spread of rumors, like fake news, in social networks. Unlikely in CIMP, there are different strategies to contrast the rumor: blocking nodes or edges or generating a positive cascade that contrasts the diffusion of the rumor. The propagation of information in the network plays a central role in these problems and determines which nodes get active and how the information is forwarded. Almost all the previous works assume that once a node gets active it remains active and therefore, they adopt the so-called progressive diffusion model. For the rumor containment problems, such simplistic diffusion models may be unrealistic, and semi-progressive or non-progressive models would be more suitable. Although these more complex diffusion models may lead to better strategies to contrast rumors, they have attracted little attention in the literature. The overall goal of my research is to study and develop discrete optimization models and methods for extended cases and variants of rumor containment problems, including semi-progressive and non-progressive diffusion models.