Iterative solvers for linear systems with an early stopping rule provide a fast and efficient way of solving ill-posed inverse problems. The regularization by stopping the iterations, however, does not translate directly into the language of Bayesian priors. In this talk, the stopping rules are revisited in the Bayesian context and the question how to modify them to correspond to prior information is addressed.

This presentation is part of Minisymposium “MS54 - Hybrid Approaches that Combine Deterministic and Statistical Regularization for Applied Inverse Problems (4 parts)”
organized by: Cristiana Sebu (University of Malta) , Taufiquar Khan (Clemson University) .

Authors:

Erkki Somersalo (Case Western Reserve University)

Keywords:

bayesian methods, inverse problems