Online supporting information.

S1 Algorithm. Simulation-based offline learning. S2 Algorithm. Simulation-based online learning. S3 Algorithm. Simulation-based hybrid learning. S1 Table. Disease model priors. S2 Table. Intervention model priors. S3 Table. Observation model priors. S4 Table. Validation SIR model priors. S5 Table. Periodic SIR model priors. S1 Fig. Model-based predictions with non-identifiable parameters. S2 Fig. Calibration of OutbreakFlow trained with additional non-identifiable parameters. S3 Fig. Bivariate posteriors from real data. S4 Fig. Calibration of OutbreakFlow trained on the periodic SIR model. S5 Fig. Model-based predictions on 16 simulated time-series from the periodic SIR model. S6 Fig. Calibration of OutbreakFlow trained for inference on entire Germany. S7 Fig. Calibration of OutbreakFlow without a convolutional filtering network. S8 Fig. Model-based predictions using OutbreakFlow without a convolutional filtering network. S9 Fig. Calibration of OutbreakFlow without a recurrent summary network. S10 Fig. Model-based predictions using OutbreakFlow without a recurrent summary network. S11 Fig. Calibration of OutbreakFlow without a latent carrier compartment in the epidemiological model. S12 Fig. Model-based predictions using an epidemiological model without a carrier compartment. S13 Fig. Calibration of OutbreakFlow trained for inference on the German federal states. S14 Fig. Model predictions of new cases for each German federal state. S15 Fig. Model predictions of cumulative Covid-19 deaths (derived from predicted new cases) for each German federal state. S16 Fig. Marginal parameter posteriors from data available for the German federal state Baden-Württemberg. S17 Fig. Marginal parameter posteriors from data available for the German federal state Bavaria. S18 Fig. Marginal parameter posteriors from data available for the German federal state Berlin. S19 Fig. Marginal parameter posteriors from data available for the German federal state Brandenburg. S20 Fig. Marginal parameter posteriors from data available for the German federal state Bremen. S21 Fig. Marginal parameter posteriors from data available for the German federal state Hamburg. S22 Fig. Marginal parameter posteriors from data available for the German federal state Hesse. S23 Fig. Marginal parameter posteriors from data available for the German federal state Saxony. S24 Fig. Marginal parameter posteriors from data available for the German federal state Mecklenburg-Western Pomerania. S25 Fig. Marginal parameter posteriors from data available for the German federal state North Rhine-Westphalia. S26 Fig. Marginal parameter posteriors from data available for the German federal state Rhineland-Palatinate. S27 Fig. Marginal parameter posteriors from data available for the German federal state Saarland. S28 Fig. Marginal parameter posteriors from data available for the German federal state Saxony-Anhalt. S29 Fig. Marginal parameter posteriors from data available for the German federal state Saxony. S30 Fig. Marginal parameter posteriors from data available for the German federal state Schleswig-Holstein. S31 Fig. Marginal parameter posteriors from data available for the German federal state Thuringia.

(PDF)