PhD Defense: Scaling Agent-based Epidemic Diffusion on HPC Clusters

Over the course of the COVID-19 pandemic, policymakers at various levels of government have harnessed a wide variety of modeling approaches to inform their decisions. In particular, agent-based models (ABMs), in which the behavior of individual agents within the modeled population is simulated directly, have proved effective at capturing the impacts of a range of public health interventions. However, ABMs are orders of magnitude more computationally intensive to run than conventional compartmental models, which use systems of ordinary differential equations. As a result, existing ABMs in this space are generally either (1) complex, small simulations (at most around a million agents) with a focus on epidemiological results rather than computational efficiency, or (2) simpler, large models where much of the complexity lies in the underlying datasets and the behavioral model of agents remains relatively simple (e.g. a sequence of top-down interventions determines behavior). In addition, the latter often have difficulty scaling efficiently to large core counts, and generally experiences significant performance degradation when simulating more complex interventions (e.g. contact tracing). With these limitations of existing simulations in mind, we set out to enable ABMs of infectious disease spread to efficiently scale to large populations and core counts while efficiently modeling a combination of top-down and bottom-up behaviors that are both complex and dynamic. There are two main directions involved in this work: (1) increasing the scalability of ABMs for large populations and (2) introducing more complex behavioral models into large-scale ABMs, particularly ones which dynamically evolve with the spread of the simulated disease.