Pandemic situation awareness: towards trustworthy model-based epidemiological analysis

This proposal aims to develop computational approaches to address modeling challenges observed during the Covid-19 pandemic. Specifically, the project seeks to enhance data integration and management of uncertainties in epidemiological models.

The research will be conducted within an epidemiological simulation framework and Bayesian methods lie at the core of the project.

Concretely, we willDevelop a neural network-based workflow which, for a considered computational model, can produce effective Bayesian priors constrained by a maximum entropy condition.Design and implement a family of epidemiological models, offering flexible control of the modeling granularity, and with built-in support for approximate likelihoods based on state estimation techniques.Establish computable estimates for the impact of incomplete models and the use of approximate likelihoods, hence allowing for a posteriori diagnostics and corrections.The societal impact of this research will be in increased pandemic preparedness, decision situation awareness, and agile epidemiological response.

The high-level purpose is to advance efficient algorithms and workflows for combining computational models with uncertain data and, considering the targeted application, specifically with high requirements on trustworthiness.The project is scheduled for 4-years involving myself, a PostDoc to be recruited within the first year, and an already recruited PhD student.