Data-driven approaches for predicting spread of infectious diseases through DINNs: Disease Informed Neural Networks
This work addresses forecasting infectious disease spread for public health planning, but it is incremental as it builds on existing physics-informed neural network methods.
The authors tackled predicting infectious disease spread by introducing Disease Informed Neural Networks (DINNs), which adapt physics-informed neural networks to SIR compartmental models and other disease models, showing it can learn dynamics and forecast progression from real-world data across eleven diseases.
In this work, we present an approach called Disease Informed Neural Networks (DINNs) that can be employed to effectively predict the spread of infectious diseases. This approach builds on a successful physics informed neural network approaches that have been applied to a variety of applications that can be modeled by linear and non-linear ordinary and partial differential equations. Specifically, we build on the application of PINNs to SIR compartmental models and expand it a scaffolded family of mathematical models describing various infectious diseases. We show how the neural networks are capable of learning how diseases spread, forecasting their progression, and finding their unique parameters (e.g. death rate). To demonstrate the robustness and efficacy of DINNs, we apply the approach to eleven highly infectious diseases that have been modeled in increasing levels of complexity. Our computational experiments suggest that DINNs is a reliable candidate for effectively learn about the dynamics of spread and forecast its progression into the future from available real-world data.