TechRxiv
L1o2uniform0728FinalTaoZhang.pdf (696.82 kB)

A Unified Study on L1 over L2 Minimization

Download (696.82 kB)
preprint
posted on 05.08.2021, 10:49 by Min TaoMin Tao, Xiao-Ping Zhang
In this paper, we carry out a unified study for L_1 over L_2 sparsity promoting models, which are widely used in the regime of coherent dictionaries for recovering sparse nonnegative/arbitrary signal. First, we provide the exact recovery condition on both the constrained and the unconstrained models for a broad set of signals. Next, we prove the solution existence of these L_{1}/L_{2} models under the assumption that the null space of the measurement matrix satisfies the $s$-spherical section property. Then by deriving an analytical solution for the proximal operator of the L_{1} / L_{2} with nonnegative constraint, we develop a new alternating direction method of multipliers based method (ADMM$_p^+$) to solve the unconstrained model. We establish its global convergence to a d-stationary solution (sharpest stationary) and its local linear convergence under certain conditions. Numerical simulations on two specific applications confirm the superior of ADMM$_p^+$ over the state-of-the-art methods in sparse recovery. ADMM$_p^+$ reduces computational time by about $95\%\sim99\%$ while achieving a much higher accuracy compared to commonly used scaled gradient projection method for wavelength misalignment problem.

Funding

National Key Research and Development Program of China (2018AAA0101100)

the Natural Science Foundation of China (No. 11971228)

the Jiangsu Provincial National Natural Science Foundation of China (No. BK20181257)

the Natural Sciences and Engineering Research Council of Canada (NSERC), Grant No. RGPIN-2020-04661.

History

Email Address of Submitting Author

taom@nju.edu.cn

ORCID of Submitting Author

0000-0003-4750-2959

Submitting Author's Institution

Nanjing University

Submitting Author's Country

China