Dr Briol awarded EPSRC grant on ‘Robust Foundations for Bayesian Inference’

Congratulations to Dr Briol for his recent success in getting an EPSRC small grant in the mathematical sciences on the topic of ‘Robust Foundations for Bayesian Inference’. The grant will support Dr Briol (PI) and Dr Knoblauch (as co-I) for a year whilst they investigate the problem of model misspecification in Bayesian statistics and machine learning.

Model misspecification is the problem of making certain modelling assumptions which turn out to be incorrect in practice. This can for example happen in scientific applications when models that scientists have developed are missing some key bits of the physics, chemistry or biology of a problem, or when making some measurements which turn out to be subject to unexpected errors. These incorrect modelling assumptions can have significant implications when using these models for prediction, and lead to misleading conclusions.

This EPSRC grant will support Dr Briol and Dr Knoblauch in their investigation of cases where Bayesian models are ‘robust’ to model misspecification, meaning they are able to make good predictions despite broken assumptions, and in their development of new tools for measuring Bayesian robustness in practice.