OutbreakFlow: Model-based Bayesian inference of disease outbreak dynamics with invertible neural networks and its application to the COVID-19 pandemics in Germany
Fig 4
(a) Posterior predictions and forecasts of new cases obtained by inferring model parameters from epidemiological data available for reported infected, assumed recovered and deaths by Covid-19 for the entire Germany. Cases to the left of the vertical dashed line were used for posterior checking (model training) and cases to the right for posterior forecasts (predictions) on unseen data; (c) Marginal posteriors of all 34 model parameters inferred from data for the entire Germany alongside median and MAP summary statistics. Gray lines depict prior distributions for comparison with the posteriors. Vertical dashed lines indicate posterior medians.