Contagion-Preserving Network Sparsifiers: Exploring Epidemic Edge Importance Utilizing Effective Resistance

Abstract: Network epidemiology has become a vital tool in understanding the effects of high-degree vertices, geographic and demographic communities, and other inhomogeneities in social structure on the spread of disease. However, many networks derived from modern datasets are quite dense, such as mobility networks where each location has links to a large number of potential destinations. One way to reduce the computational effort of simulating epidemics on these networks is sparsification, where we select a representative subset of edges based on some measure of their importance. Recently an approach was proposed using an algorithm based on the effective resistance of the edges. We explore how effective resistance is correlated with the probability that an edge transmits disease in the SI model. We find that in some cases these two notions of edge importance are well correlated, making effective resistance a computationally efficient proxy for the importance of an edge to epidemic spread. In other cases, the correlation is weaker, and we discuss situations in which effective resistance is not a good proxy for epidemic importance.