Location
MC 6460
Speaker
Dr. Woldegebriel Assefa Woldegerima, Assistant Professor of the Department of Mathematics and Statistics, York University
Title
Modeling and Predicting Infectious Diseases Dynamics through Epidemiology-Informed Neural Networks (EINNs)
Abstract
Infectious disease modeling and prediction remains a crucial research area as new variants continue to (re)emerge. The integration of mechanistic modeling and machine learning has the potential to revolutionize the way we understand complex biological systems. Particularly, the development and application of Informed Neural Networks (INNs), a class of hybrid models that embed the domain-specific knowledge, such as differential equations, into the neural network architecture has attracts many researchers recently. Epidemiology-Informed Neural Networks (EINNs) incorporate domain-specific knowledge from disease dynamics (e.g., differential equation models, compartmental models like SIR, vector-host interactions, etc.) into their architecture, loss functions, or training process. The loss function that is minimized during training is the combined loss of the data and the DEs residuals. This method not only provides enhanced data accuracy but also has the domain knowledge to predict how the disease evolves. It helps to improve learning, prediction accuracy, interpretability, and parameter estimation, particularly in scenarios where data is sparse or noisy. In this talk I will quickly introduction the foundations of EINNs. I will then present some results from our study that we e trained an EINN on synthetic data derived from an SI-SIR model designed for Avian influenza and shows the model’s accuracy in predicting extinction and persistence conditions. In the method, a twelve hidden layer model was constructed with sixty-four neurons per layer and ReLU activation function was used. The network is trained to predict the time evolution of five state variables for birds and humans over 50,000 epochs. The overall loss minimized to 0.000006, characterized by a combination of data and physics losses, enabling the EINN to follow the differential equations describing the disease progression.
(If time permits), I will also briefly present some of my recent publication outputs on in-host modelling of malaria parasites, data-drive modelling of immune responses and treatment Mpox in-host dynamic, and other population compartmental transmission models of Mpox with behavioural change, and risk-structured classes.