1: \begin{abstract}

2: Ill-posed inverse problems are ubiquitous in applications. Understanding

3: of algorithms for their solution has been greatly enhanced by a deep

4: understanding of the linear inverse problem. In the applied communities

5: ensemble-based filtering methods have recently been used to solve

6: inverse problems by introducing an artificial dynamical system. This

7: opens up the possibility of using a range of other filtering methods,

8: such as 3DVAR and Kalman based methods, to solve inverse problems,

9: again by introducing an artificial dynamical system. The aim of this paper

10: is to analyze such methods in the context of the ill-posed linear inverse

11: problem.

12: 

13: 

14: Statistical linear inverse problems are studied in the sense that the

15: observational noise is assumed to be derived via realization of

16: a Gaussian random variable.  We investigate the asymptotic behavior of filter

17: based methods for these inverse problems. Rigorous

18: convergence rates are established for 3DVAR and for the Kalman filters,

19: including minimax rates in some instances. Blowup of 3DVAR and a variant

20: of its basic form is also presented,

21: and optimality of the Kalman filter is discussed.

22: These analyses reveal a close connection between (iterative) regularization

23: schemes in deterministic inverse problems and filter based methods

24: in data assimilation. Numerical experiments are presented to illustrate

25: the theory.

26: \end{abstract}

27: