Figure 3

Prediction performance of each method in terms of epidemic impact, for each data set and for various β, at fixed \(\mu =0.055\). The results at fixed β and for various μ are shown as a Supplementary Figure. The top and bottom row show respectively the cumulative discrepancy between the real and predicted epidemic curves \(\Delta _{I}\) and the discrepancy between the final epidemic sizes \(\Delta _{\mathrm{size}}\). Error bars give the standard deviation computed over the realizations of the SIR process and the embeddings