Augmented L1 and Nuclear-Norm Models with a Globally Linearly Convergent Algorithm
Abstract
This paper studies the long-existing idea of adding a nice smooth function to "smooth" a non-differentiable objective function in the context of sparse optimization, in particular, the minimization of , where is a vector, as well as the minimization of , where is a matrix and and are the nuclear and Frobenius norms of , respectively. We show that they can efficiently recover sparse vectors and low-rank matrices. In particular, they enjoy exact and stable recovery guarantees similar to those known for minimizing and under the conditions on the sensing operator such as its null-space property, restricted isometry property, spherical section property, or RIPless property. To recover a (nearly) sparse vector , minimizing returns (nearly) the same solution as minimizing almost whenever . The same relation also holds between minimizing and minimizing for recovering a (nearly) low-rank matrix , if . Furthermore, we show that the linearized Bregman algorithm for minimizing subject to enjoys global linear convergence as long as a nonzero solution exists, and we give an explicit rate of convergence. The convergence property does not require a solution solution or any properties on . To our knowledge, this is the best known global convergence result for first-order sparse optimization algorithms.
Cite
@article{arxiv.1201.4615,
title = {Augmented L1 and Nuclear-Norm Models with a Globally Linearly Convergent Algorithm},
author = {Ming-Jun Lai and Wotao Yin},
journal= {arXiv preprint arXiv:1201.4615},
year = {2015}
}
Comments
arXiv admin note: text overlap with arXiv:1207.5326 by other authors