6533b851fe1ef96bd12a89d3

RESEARCH PRODUCT

Impact of the community structure on the dynamics of complex networks

subject

description

Networks are everywhere. We encounter them daily in our lives, through our social interactions, how we come up with decisions in our brain, to having phone calls, conducting financial transactions, and traveling from one place to another. Individual actions are influenced by their environment, which is, in turn, influenced by the network's topology. Notably, individuals may change their actions, ideas, or opinions to conform to the aspirations of a particular social group. In the same vein, the spread of a virus can take a certain course if the network's structure induces specific pathways for expansion. In such scenarios, communities substantially impact the evolution of the dynamics. They can hinder or enhance diffusion flow depending on where diffusion originates. Nodes within and between communities are responsible for initiating the dynamic diffusion flow in networks, while influential nodes can play a crucial role in boosting diffusion. The significance of comprehending the community structure of a network and its impact on the underlying dynamics, initiated by the nodes, is accentuated by many real-world scenarios. In this thesis, we study the interplay between dynamic models, influential nodes, the process of identifying them, and the network's topology. First, we investigate how the output of various dynamic models is influenced by the network topology, with seed nodes ranked using community-aware centrality measures. Studying this problem can provide insights into how diffusion spreads and identify constraints that limit the effectiveness of utilizing dynamic scenarios in practical situations, such as promoting viral marketing or combating false information. Second, we tackle the problem of influence redundancy and propose a new ranking scheme to naturally selects distant nodes to expand any diffusion phenomena. By tackling this problem through the proposed ranking scheme, diffusion ought to be maximized, independent of the network type. This renders a powerful tool suitable for researchers aiming to maximize diffusion in many applications. Third, researchers mainly focus on identifying influential nodes in networks with a non-overlapping community structure, while many networks have an overlapping community structure. Moreover, the measures developed for networks with an overlapping community structure are inflexible to missing or varying information. Therefore, we propose a flexible framework that identifies influential nodes in networks with incomplete, complete, fuzzy, or crisp overlapping information about the nodes. This framework allows researchers to incorporate various information about overlaps and customize it to different circumstances and information availability.