Bayesian uncertainty quantification for epidemic spread on networks

While there exist a number of mathematical approaches to modeling the spread of disease on a network, analyzing such systems in the presence of uncertainty introduces significant complexity. In scenarios where system parameters must be inferred from limited observations, general approaches to uncertainty quantification can generate approximate distributions of the unknown parameters, but these methods often become computationally expensive if the underlying disease model is complex. In this paper, we apply the recent massively parallelizable Bayesian uncertainty quantification framework $\Pi4U$ to a model of a disease spreading on a network of communities, showing that the method can accurately and tractably recover system parameters and select optimal models in this setting.