DEpiABS: Differentiable Epidemic Agent-Based Simulator

Researchers have unveiled DEpiABS, a fully differentiable agent-based simulator that significantly improves epidemic forecasting accuracy. The model reduces average forecasting deviation from 0.97 to 0.92 on COVID-19 mortality data and from 0.41 to 0.32 on influenza data. Crucially, it achieves this 5-9% improvement without requiring auxiliary data, using a novel scaling method to simulate large populations efficiently while maintaining individual-level detail on health, behavior, and viral mutation.