Processus de contagion sur réseaux complexes au-delà des interactions dyadiques

Authors: St-Onge, Guillaume
Advisor: Allard, AntoineHébert-Dufresne, Laurent
Abstract: After almost two years into the COVID-19 pandemic, it is clear that a better understanding of contagion processes, their evolution, and the impact of control measures is essential to reduce their burden on society. The theoretical framework for the modeling of contagion is quite general. It can describe the spread of pathogens causing diseases (viruses, bacteria, parasites, etc.), but also the spread of rumors and disinformation. Irrespective of the nature of the underlying process, the contagion evolves through local interactions between the individuals. Consequently, the complex social structure of populations, which is neither perfectly ordered nor completely random, plays a fundamental role in shaping spreading. In this thesis, we study contagion processes on networks where individuals and the interaction between them are represented by nodes and edges respectively. We use a theoretical approach based on statistical physics and nonlinear dynamics. We focus on higher-order networks, putting group interactions beyond pairwise interactions at the forefront. More than a mere mathematical generalization, we find this perspective is paramount to obtain a complete picture of the phenomenology of contagion dynamics. We demonstrate the importance of group interactions through three principal results. First, we characterize a mesoscopic localization phenomenon where the contagion thrives only in large groups for certain types of heterogeneous structure. This phenomenon significantly affects the results of interventions like the cancelation of events larger than a critical size, similar to the measures being used to limit the spreading of COVID-19. Second, we study a model where individuals must accumulate a minimal infective dose to become infected. We show that a higher-order structure and heterogeneous exposure induce a universal nonlinear infection probability. The epidemic size can then grow super-exponentially with time. Finally, with a more in-depth analysis of nonlinear contagions, we show that groups can be more influential than hubs (super-connected individuals) to maximize the early spread of an epidemic.
Document Type: Thèse de doctorat
Issue Date: 2022
Open Access Date: 21 March 2022
Grantor: Université Laval
Collection:Thèses et mémoires

Files in this item:
Description SizeFormat 
37837.pdf4.36 MBAdobe PDFThumbnail
All documents in CorpusUL are protected by Copyright Act of Canada.