Title: “Bayesian inference with detailed physical models”

Speaker: Dr. Youssef Marzouk

Date/Time: Thursday, June 22, 2006, 2:00 pm (MST)

Location: Building 980, Room 95 (Sandia NM)

Brief Abstract: Bayesian statistics provides a foundation for inference from noisy data and stochastic forward models, a natural mechanism for incorporating prior information, and a quantitative assessment of uncertainty in the inferred results. Inverse problems representing indirect estimation of model parameters, inputs, or structural components can be fruitfully cast in this framework. The computationally intensive forward models arising in many physical applications, however, can render a Bayesian approach prohibitive.

We develop new computational tools for Bayesian inference in this context, showing strong connections between uncertainty propagation and Bayesian inference. In particular, we present a stochastic spectral formulation that dramatically accelerates the Bayesian solution of inverse problems. These methods are evaluated on transient diffusion problems arising in contaminant source inversion. We also discuss dimensionality reduction in inverse problems, again using stochastic spectral approaches.

CSRI POC: Bart van Bloemen Waanders, (505) 284-6746