Speaker: Elizaveta Semenova, ML Researcher, Oxford & Imperial College London
Abstract
Disease mapping is an important surveillance tool that enables researchers and public health officials to analyse the spatial distribution of a disease, identify its geographical patterns, and plan interventions. Disease mapping has been used for analysis and communication in public health since the 19th century. Contemporary models for disease mapping rely on hierarchical models capturing data structure and Bayesian inference for uncertainty quantification. While being well established, this methodology is falling behind compared to other modern computational domains. For example, generative AI is creating an undeniable impact in several scientific fields, while its potential in epidemiology and public health remains largely untapped. In this talk I will explain how deep generative models can help spatial statistics, and epidemiology more generally. The talk will begin with an introduction to spatial statistics and disease mapping. After that I will present the idea of learning priors with deep generative models and explain how such priors can enable faster and more efficient Bayesian inference with Markov Chain Monte Carlo (MCMC) algorithms. At the end of the talk, I will present another flavour of spatial statistics — modelling on graphs, and outline ongoing research directions using geometric learning.