Modeling COVID-19 Dynamics in German States Using Physics-Informed Neural Networks

Phillip Rothenbeck, Sai Karthikeya Vemuri, Niklas Penzel, Joachim Denzler

Keywords: Compartmental Models, COVID-19, Physics Informed Neural Networks, Pandemic Modeling

TL;DR: We use PINNs to analyze COVID-19 dynamics in German federal states by solving the inverse problem of the SIR model.

Abstract:

We present a three-year, state-level analysis of COVID-19 in Germany using a simple SIR model coupled with Physics-Informed Neural Networks (PINNs), trained on data from the Robert Koch Institute.
For each of the 16 German federal states, we use the PINN framework to estimate transmission and recovery rates $(\beta, \alpha)$ and the time-dependent reproduction number $\mathcal{R}_t$.
Our results showcase significant regional heterogeneity and an inverse relationship between vaccination uptake and both $\beta$ and peak $\mathcal{R}_t$ numbers.
Furthermore, we observe that the inferred progression of $\mathcal{R}_t$ aligns with the major phases of the pandemic, including the Omicron peak, followed by stabilization at or below the epidemic threshold of 1.0 by mid-2022.
These findings demonstrate the utility of PINNs for localized, long-term epidemiological modeling and evaluating regional policy impacts.

Email Sharing: We authorize the sharing of all author emails with Program Chairs.

Data Release: We authorize the release of our submission and author names to the public in the event of acceptance.

Submission Number: 27