Application of some deterministic techniques to Bayesian inference

Quantifying the uncertainty in computational simulations is one of the central challenges confronting the field of computational science and engineering today. The uncertainty quantification of inverse problems is neatly addressed in the Bayesian framework, where instead of seeking one unique minimiser of a regularised misfit functional, the entire posterior probability distribution is to be characterised. In this talk I review the deep connection between deterministic PDE-constrained optimisation techniques and Bayesian inference for inverse problems, discuss some recent advances made in the Bayesian viewpoint by adapting deterministic techniques, and mention directions for future research.