Dimension reduction for Bayesian inference of large-scale systems

Authors

Tiangang Cui, Massachusetts Institute of Technology

Document Type

Presentation Abstract

Presentation Date

11-9-2015

Abstract

Algorithmic scalability to high dimensional parameters and computational efficiency of numerical solvers are two significant challenges in large-scale Bayesian inversion. Here we will explore the intrinsic dimensionality in both state space and parameter space of inverse problems by analyzing the interplay between noisy data, ill-posed forward model and smoothing prior. The resulting reduced subspaces naturally lead to a scalable and fast model reduction framework for solving large-scale inverse problems with high dimensional parameters.

Additional Details

Monday, November 9, 2015 at 3:10 p.m. in Math 103
Refreshments at 4:00 p.m. in Math Lounge 109

Recommended Citation

Cui, Tiangang, "Dimension reduction for Bayesian inference of large-scale systems" (2015). Colloquia of the Department of Mathematical Sciences. 485.
https://scholarworks.umt.edu/mathcolloquia/485