Accurate medium- to long-range epidemic forecasting is a hard problem. It requires extrapolating high-dimensional, heterogeneous, and nonstationary dynamics from partial observations. A new arXiv paper introduces a hybrid model designed for this task.
Hybrid models often leave implicit which observations inform which parts of the dynamics. The proposed approach uses Lagrangian-Eulerian graph neural ODEs to make this relationship explicit. This is intended to improve the model's ability to forecast across multiple spatial and temporal scales.
The abstract does not include experimental results, so the claims are about the model's design and motivation. The paper is available on arXiv under identifier 2610.21223.