Machine-Learned Epidemiology
Abstract: Work in computational epidemiology to date has been limited by coarseness and lack of timeliness of observational data. Most existing models are based on hand-curated statistics that are often delayed, expensive to collect, and cover only limited jurisdictions. Our goal is to lift the state of the art in epidemiology to a new qualitative state, where real-time health predictions become feasible and actionable. We do this at scale by applying machine learning and secure aggregation to online data to infer what likely contributed to the contagion. In this talk, I will sample current projects at Google focusing on privacy-first epidemiology research and recent publications (e.g., [1], [2], [3], [4], [5]).