A fast and scalable computational framework for goal-oriented linear Bayesian optimal experimental design: Application to optimal sensor placement

Optimal experimental design (OED) is a principled framework for maximizing information gained from limited data in inverse problems. Unfortunately, conventional methods for OED are prohibitive when applied to expensive models with high-dimensional parameters, as we target here. We develop a fast and scalable computational framework for goal-oriented OED of large-scale Bayesian linear inverse problems that finds sensor locations to maximize the expected information gain (EIG) for a predicted quantity of interest. By employing low-rank approximations of appropriate operators, an online-offline decomposition, and a new swapping greedy algorithm, we are able to maximize EIG at a cost measured in model solutions that is independent of the problem dimensions. We demonstrate the efficiency, accuracy, and both data- and parameter-dimension independence of the proposed algorithm for a contaminant transport inverse problem with infinite-dimensional parameter field.