Skip to main content

"Beyond MCMC -- scalable and approximate Bayesian inference for computational statistics in global health"

Supervisor

Suitable for

MSc in Advanced Computer Science
Computer Science and Philosophy, Part C
Mathematics and Computer Science, Part C
Computer Science, Part C
Computer Science, Part B

Abstract

Co-supervisor: Dr Swapnil Mishra (https://s-mishra.github.io/)

In applied work, especially disease modeling, we have reached the limits of what standard MCMC samplers can solve in a reasonable amount of computational time. We will investigate a variety of recently proposed inference schemes, from variational Bayes to deep learning ensemble models to parallel implementations of Sequential Monte Carlo, applying them to popular models in biostatistics, especially multilevel / hierarchical models. The goal will not be just to figure out "what works" but to understand the shortcomings of existing tools, and come up with guidance for practitioners. Co-supervisor