Sparse Recovery
Overview
Sparse recovery asks how to identify a small set of nonzero coefficients and estimate their values from incomplete or noisy observations, a problem central to compressed sensing, imaging, and high-dimensional regression. Common approaches include convex ℓ1 regularization, greedy and hard-thresholding algorithms, Bayesian and nonconvex penalties, and learned or unrolled methods that incorporate structure such as groups, blocks, or dictionaries. Theory uses tools such as restricted isometry, mutual coherence, and geometric complexity to establish recovery conditions, while newer work extends guarantees to online, nonlinear, and adversarially corrupted settings. A continuing challenge is balancing statistical accuracy and robustness against computational cost, especially when measurements are highly correlated or the assumed sparse representation is imperfect.
Latest papers
77 papers11 papers in the last four weeks, up 38% on the four weeks before. 0.1% of all new papers.