Memory-induced complex contagion in spreading phenomena on networks.
Supervised by Dr. Marián Boguñá Espinal at the University of Barcelona.
June 2021, Xavier R. Hoffmann [email protected]
This repository is licensed under the Creative Commons Attribution-ShareAlike 4.0 International license.
- See hoffmann_phd_digital.pdf for the full version of my PhD thesis.
- See hoffmann_phd_short.pdf for a short version including only abstract, introduction and conclusions.
Epidemic modeling has proven to be an essential framework for the study of contagion phenomena in biological, social, and technical systems. Albeit epidemic models have evolved into powerful predictive tools, most assume memoryless agents and independent transmission channels. Nevertheless, many real-life examples are manifestly time-sensitive and show strong correlations. Moreover, recent trends in agent-based modeling support a generalized shift from edge-based descriptions toward node-centric approaches.
Here I develop an infection mechanism that is endowed with memory of past exposures and simultaneously incorporates the joint effect of multiple infectious sources. A notion of social reinforcement/inhibition arises organically, without being incorporated explicitly into the model. As a result, the concepts of non-Markovian dynamics and complex contagion become intrinsically coupled. I derive mean-field approximations for random degree-regular networks and perform extensive stochastic simulations for nonhomogeneous networks.
The analysis of the SIS model reveals a sophisticated interplay between two memory modes, displayed by a collective memory loss and the dislocation of the critical point into two phase transitions. An intermediate region emerges where the system is either excitable or bistable, exhibiting fundamentally distinct behaviors compared to the customary healthy and endemic phases. Additionally, the transition to the endemic phase becomes hybrid, showing both continuous and discontinuous properties.
These results provide renewed insights on the interaction between microscopic mechanisms and topological aspects of the underlying contact networks, and their joint effect on the properties of spreading processes. In particular, this type of modeling approach that combines memory effects and complex contagion could be suitable to describe ecological interactions between biological and social pathogens.