1: \begin{abstract}

2:  We introduce a new method to reconstruct unknown quantum states out

3: of incomplete and noisy information. 

4: The method is a linear convex optimization problem, therefore with a unique

5: minimum, which can be efficiently solved with Semidefinite Programs.

6: Numerical simulations indicate that the estimated state does not overestimate

7: purity, and neither the expectation value of  optimal entanglement witnesses.

8: The convergence properties of the method are similar to {\em compressed 

9: sensing} approaches, in the sense that, in order to reconstruct

10: low rank states,  it needs just a fraction of the effort

11:  correspondig to an informationally complete measurement.

12: 

13: \end{abstract}

14: