Project title: Applications of AI/ML to epidemiological time series
PI: David Gavaghan
Infectious diseases such as COVID-19 represent an important target for scientific modelling. Biological knowledge motivates the development of mathematical models which describe how diseases progress through a population. These models are characterized by one or more parameters, which quantitatively specify various properties of the model. In some cases, parameter values are informed by prior knowledge of the disease; frequently, however, we must infer parameter values on the basis of discrete time series data of monitored values such as the daily number of deaths or weekly number of positive tests. This inference problem raises numerous challenges which will benefit from the application of advanced techniques from AI and ML. In particular, the data underlying epidemiological inference suffers from noise and bias, motivating the development of more robustness in epidemiological models, and the extension of debiasing algorithms to improve data quality. Additionally, learning time variation in parameter values is essential for monitoring disease outbreaks, but epidemiological models demand more flexible approaches to learning time variation. These challenges are shared by other scientific fields involving parameter inference from time series data, and the lessons learned by applying AI and ML in the epidemiological setting are expected to apply directly to other problems.
