High Dimensional Bayesian Integration via Sparsely Parameterized Transport Maps and Quadrature
Friday, April 26, 2019
11AM – 12PM
In uncertainty quantification, one has to often face the problem of efficient posterior integration for computing moments of quantities of interest (QoIs) in high dimension. One method to do this is dimension-adaptive sparse quadrature. I describe efforts towards scalable dimension-adaptive sparse quadrature via constructing a sparsely parameterized push forward transport map between the prior and the posterior. Along the way, I discuss the connection between sparsity and semilattices.