1: \begin{abstract}

2: We consider nonlinear inverse problems described by operator equations

3: in Banach spaces. Assuming conditional stability of the inverse

4: problem, that is, assuming that stability holds on a closed, convex

5: subset of the domain of the operator, we introduce a novel nonlinear

6: projected steepest descent iteration and analyze its convergence to an

7: approximate solution given limited accuracy data. We proceed with

8: developing a multi-level algorithm based on a nested family of closed,

9: convex subsets on which stability holds and the stability constants

10: are ordered. Growth of the stability constants is coupled to the increase in

11: accuracy of approximation between neighboring levels to ensure that the

12: algorithm can continue from level to level until the iterate satisfies

13: a desired discrepancy criterion, after a finite number of steps.

14: \end{abstract}