1: \begin{abstract}

2: We propose a new gradient projection algorithm that compares

3: favorably with the fastest algorithms available to date for

4: $\ell_1$-constrained sparse recovery from noisy data, both in

5: the compressed sensing and inverse problem frameworks. The

6: method exploits a line-search along the feasible direction and

7: an adaptive steplength selection based on recent strategies for

8: the alternation of the well-known Barzilai-Borwein rules. The

9: convergence of the proposed approach is discussed and a

10: computational study on both well-conditioned and

11: ill-conditioned problems is carried out for performance

12: evaluations in comparison with five other algorithms proposed

13: in the literature.

14: \end{abstract}

15: