Refactoring a pandemic modelling tool for computational speed and methodological flexibility
Feb 9, 2022
3:30PM to 5:00PM
Date/Time
Date(s) - 09/02/2022
3:30 pm - 5:00 pm
Speaker: Dr. Steve Walker
Topic: Refactoring a pandemic modelling tool for computational speed and methodological flexibility
Date of Presentation: Wednesday, February 09, 2022
Location: Online
Abstract:
McMasterPandemic is a modelling tool that was rapidly developed to provide timely insights into the Covid-19 Pandemic in Canada. This tool has been used throughout the pandemic to provide forecasts and insights to the Ontario Science Advisory Table and the Public Health Agency of Canada. McMasterPandemic has been used to do interesting things, like predicting the third Covid wave in Ontario while case numbers were declining. To be relevant in policy decisions, the McMasterPandemic modelling team needed to provide forecasts at rapid speeds by volunteers with full-time jobs. As a result of these realities, the team was often forced to prioritize short-term deliverables over flexibility, ease-of-use, and computational performance. Still, the mathematical and computational foundation of McMasterPandemic has the potential to grow into a mature and general tool for forecasting infectious disease dynamics. My work on McMasterPandemic has been on refactoring it so that it is faster and more general. I will describe our performance gains to date (10-100x), the current state of a generalized modelling interface, and our future plans.