CAOS · Containment, Avalanches and Optimisation in Spreading-processes

Spreading processes, such as cascading failure of technological networks, epidemic outbreak and contagion in socialnetworks are ubiquitous and have a significant impact on the modern society. In many cases, transitions to large-scaleblackouts or global pandemics occur abruptly due to tiny changes in system parameters; epidemics and social movementscan spread out explosively, forming avalanche-like outbreaks in a very short period. The drastic transitions are beneficial formaking global impact of boosting charity campaigns and commercial advertising but can also have catastrophicconsequences when epidemics or failures spread to large part of a critical systems abruptly, without any early warningsignals. Moreover, multiple cooperative of competitive spreading processes make the picture even more complex.Previous work in this area focused on the asymptotic manifestation of the process and typically ignored the interplaybetween specific system topology and the dynamics. Methods to monitor, identify and control drastic transitions andoutbreaks in heterogeneous realistic networks are lacking. In this project, we will investigate the signs for abrupt transitionsand avalanche outbreaks both macroscopically and in specific instances, to gain insight into the conditions for their onset.We will develop methods to contain (or facilitate) the outbreaks by optimal deployment of resources, such as applyingvaccines or distributing promotion material by employing the recently developed dynamic message passing techniques fromstatistical physics. We will collaborate with British Telecom to promote product marketing and service provision by using thenew optimisation algorithms. This project will have significant impact on the scientific understanding of non-equilibriumspreading processes, provide algorithmic solutions for practical problems in specific instances, will support policy makingdecisions and offer optimal resource allocation for commercial marketing tasks.