Time-Varying Edge Removal Scheduling for Network-based Epidemic Control

Abstract

We address edge removal scheduling of networks for epidemic control. Susceptible-infected-susceptible (SIS) model is a well-known epidemic diffusion process on networks. Each node alternates between infected and susceptible states until either all infections are removed or they reach an endemic equilibrium. An intuitive approach to suppressing the epidemic diffusion is to remove edges from the network. However, permanent removal of many edges may negatively affect social activities, resulting in severe economic damage. This paper presents a time-varying edge removal method for the SIS model. Specifically, we optimize the edge removal schedules to maximize the expected probability that the epidemic is contained in local regions at every time instance. The problem is solved with a gradient-based solver. Experiments on synthetic data show that the proposed method outperforms alternative methods.