On the strong convergence of forward-backward splitting in reconstructing jointly sparse signals (Q2670979)

The authors consider the problem of reconstructing an infinite set of finite-dimensional vectors that share a common sparsity pattern, from incomplete measurements. Applications that fit this model arise in imaging, data analysis, sensor arrays, and especially in the approximation of high-dimensional parameterized PDEs. The setting differs from [\textit{I. Daubechies} et al., Commun. Pure Appl. Math. 57, No. 11, 1413--1457 (2004; Zbl 1077.65055)] and [\textit{M. Fornasier} and \textit{H. Rauhut}, SIAM J. Numer. Anal. 46, No. 2, 577--613 (2008; Zbl 1211.65066)], which seeks to recover a finite set of infinite-dimensional vectors. The problem is phrased as a convex minimization problem with mixed norm penalty \(\ell_{2,1}\) and solved using forward-backward splitting. The major theoretical contribution is the proof of strong convergence from any starting guess, without strict convexity and compactness assumptions.