Minimization based formulations of inverse problems and their regularization

TL;DR

This paper explores a unified minimization-based framework for inverse problems, including regularization techniques that do not require solving forward problems, with applications to electrical impedance tomography.

Contribution

It introduces a general variational formulation encompassing reduced, all-at-once, and Kohn-Vogelius approaches, and develops new regularization methods using constraints and penalization.

Findings

New regularization methods applicable without solving forward problems

Application of box constraints for regularization parameter choice in EIT

Unified framework covering multiple inverse problem formulations

Abstract

The conventional way of formulating inverse problems such as identification of a (possibly infinite dimensional) parameter, is via some forward operator, which is the concatenation of the observation operator with the parameter-to-state-map for the underlying model. Recently, all-at-once formulations have been considered as an alternative to this reduced formulation, avoiding the use of a parameter-to-state map, which would sometimes lead to too restrictive conditions. Here the model and the observation are considered simultaneously as one large system with the state and the parameter as unknowns. A still more general formulation of inverse problems, containing both the reduced and the all-at-once formulation, but also the well-known and highly versatile so-called variational approach (also called Kohn-Vogelius functional approach) as special cases, is to formulate the inverse problem as a minimization problem -- instead of an equation -- for the state and parameter. Regularization can be incorporated via imposing constraints and/or adding regularization terms to the objective. In this paper, after giving a motivation by formulating the electrical impedance tomography (EIT) problem by means of the classical Kohn-Vogelius functional, we will dwell on the regularization aspects for such variational formulations in an abstract setting. Indeed, combination of regularization by constraints and by penalization leads to new methods that are applicable without solving forward problems. In particular, for the EIT problem we will consider a method employing box constraints in a very natural manner to incorporate the discrepancy principle for regularization parameter choice as well as a priori information on the searched for conductivity.