1: \begin{abstract}

2: Subspace recycling techniques have been used quite successfully for the acceleration of iterative methods for solving large-scale linear systems.  These methods often work by augmenting a solution subspace generated iteratively by a known

3: algorithm with a fixed subspace of vectors which are ``useful'' for solving the problem.  Often, this has the effect of inducing a

4: projected version of the original linear system to which the known iterative method is then applied, and this projection can

5: act as a deflation preconditioner, accelerating convergence.

6: Most often, these methods have been applied for the solution of well-posed problems.  However,

7: they have also begun to be considered for the solution of ill-posed problems.

8: 

9: In this paper, we consider subspace augmentation-type iterative schemes applied to linear ill-posed problems in a continuous

10: Hilbert space setting, based on a recently developed framework describing these methods.

11: We show that under suitable assumptions, a recycling method satisfies the formal definition

12: of a regularization, as long as the underlying scheme is itself a regularization.  We then develop an augmented subspace

13: version of the gradient descent method and demonstrate its effectiveness, both on an academic Gaussian blur model and on

14: problems arising from the adaptive optics community for the resolution of large sky images by ground-based extremely large

15: telescopes.

16: \end{abstract}

17: