Deep Gaussian processes: theory and applications

Abstract

Deep Gaussian processes have proved remarkably successful as a tool for various statistical inference tasks. This success relates in part to the flexibility of these processes and their ability to capture complex, non-stationary behaviours. In this talk, we will introduce the general framework of deep Gaussian processes, in which many examples can be constructed, and demonstrate their superiority in inverse problems including computational imaging and regression. We will discuss recent algorithmic developments for efficient sampling, as well as recent theoretical results which give crucial insight into the behaviour of the methodology.

Our Speaker

Aretha Teckentrup is Reader in the Mathematics of Data Science at the University of Edinburgh, working in Uncertainty Quantification and Bayesian inference. She obtained her PhD at the University of Bath in 2013, under supervision of Robert Scheichl in the numerical analysis group. Prior to joining the University of Edinburgh in 2016, she held postdoctoral positions at Florida State University and the University of Warwick. She is currently on the leadership team of the Prob_AI hub, focussing on mathematical and statistical foundation of probabilistic AI. Her work has been recognised by prizes including the SIAM UQ Early Career Prize, IMA Leslie Fox prize (second prize) and LMS Whitehead prize.