Iterative reweighted minimization methods for lp regularized unconstrained nonlinear programming

Research output: Contribution to journalArticlepeer-review

49 Scopus citations

Abstract

In this paper we study general lp regularized unconstrained minimization problems. In particular, we derive lower bounds for nonzero entries of the first- and second-order stationary points and hence also of local minimizers of the lp minimization problems. We extend some existing iterative reweighted l1 (IRL1) and l2(IRL2) minimization methods to solve these problems and propose new variants for them in which each subproblem has a closed-form solution. Also, we provide a unified convergence analysis for these methods. In addition, we propose a novel Lipschitz continuous ϵ-approximation to ‖x‖pp. Using this result, we develop new IRL1 methods for the lp minimization problems and show that any accumulation point of the sequence generated by these methods is a first-order stationary point, provided that the approximation parameter ϵ is below a computable threshold value. This is a remarkable result since all existing iterative reweighted minimization methods require that ϵ be dynamically updated and approach zero. Our computational results demonstrate that the new IRL1 method and the new variants generally outperform the existing IRL1 methods (Chen and Zhou in 2012; Foucart and Lai in Appl Comput Harmon Anal 26:395–407, 2009).

Original languageEnglish (US)
Pages (from-to)277-307
Number of pages31
JournalMathematical Programming
Volume147
Issue number1-2
DOIs
StatePublished - Oct 2013
Externally publishedYes

Bibliographical note

Publisher Copyright:
© 2013, Springer-Verlag Berlin Heidelberg and Mathematical Optimization Society.

Keywords

  • Iterative reweighted l1 minimization
  • Iterative reweighted l2 minimization
  • lp Minimization

Fingerprint Dive into the research topics of 'Iterative reweighted minimization methods for lp regularized unconstrained nonlinear programming'. Together they form a unique fingerprint.

Cite this