1: \begin{abstract}

2: We consider the problem of  recovering an unknown effectively \red{$(s_1,s_2)$-}sparse low-rank-$R$ matrix $X$ \red{with possibly non-orthogonal  rank-$1$ decomposition} from incomplete and inaccurate linear measurements of the form $y = \mathcal A (X) + \eta$, \red{where $\eta$ is an {\it ineliminable} noise\footnote{With ``ineliminable'' we mean that the focus of the paper is not on recovery from exact $y = \mathcal A (X)$ measurements,  rather on the stable recovery under severe measurement noise.}.}

3: %More specifically,} we aim at recovering low-rank-$R$ matrices with effectively $(s_1,s_2)$-sparse non-orthogonal rank-$1$ decompositions. 

4: We \red{first derive} an optimization formulation for matrix recovery under the considered model and propose a novel algorithm, called {\it \textbf{A}lternating \textbf{T}ikhonov regularization and \textbf{Las}so} (A-T-LA$\rm{S}_{2,1}$), to solve it. The algorithm is based on a multi-penalty regularization, which is able to leverage both structures (low-rankness and sparsity) simultaneously. 

5: The algorithm is a fast first order method, and straightforward to implement. We prove global convergence for {\it any} linear measurement model to stationary points and local convergence to global minimizers. 

6: %Global minimizers balance effective sparsity of their rank-$1$ decompositions and the fidelity to data, up to noise level. 

7: By adapting the concept of restricted isometry property from compressed sensing to our novel model class, we prove error bounds between global minimizers and ground truth, up to noise level, from a number of subgaussian measurements scaling as $R(s_1+s_2)$, up to log-factors in the dimension, and relative-to-diameter distortion. Simulation results demonstrate both the accuracy and efficacy of the algorithm, as well as its superiority to the state-of-the-art algorithms in strong noise regimes and for matrices, whose singular vectors do not possess exact (joint-) sparse support.

8: \end{abstract}

9: