Application of Physics Informed Neural Networks to understand the influence of human behavior in epidemiological modeling

Ponente(s): Alonso Gabriel Ogueda Oliva, Dr. Padmanabhan Seshaiyer

In the post-pandemic era, there have been growing efforts to develop transformative research aimed at incorporating human social, behavioral, and economic interactions in mathematical epidemiological models. However, adding this increases the complexity of the associated compartmental models and makes it challenging to estimate model parameters efficiently. In this work, we introduce and employ a novel Physics Informed Neural Networks (PINNs) approach which is extremely useful in estimating critical parameters and helps in evaluating the predictive capability of compartmental models used in epidemiology. We demonstrate how PINNS can be used to infer solutions to the governing differential equations that model a system, and estimate optimal parameters using data-driven discovery of the equations. We demonstrate through some benchmark computational experiments that our methodology is robust and a reliable candidate for parameter identification that can help to answer questions about behavioral responses to social isolation during the COVID-19 pandemic.