Suppr超能文献

一种用于非凸正则化优化问题的通用迭代收缩与阈值算法。

A General Iterative Shrinkage and Thresholding Algorithm for Non-convex Regularized Optimization Problems.

作者信息

Gong Pinghua, Zhang Changshui, Lu Zhaosong, Huang Jianhua Z, Ye Jieping

机构信息

State Key Laboratory on Intelligent Technology and Systems, Tsinghua National Laboratory for Information Science and Technology (TNList), Department of Automation, Tsinghua University, Beijing 100084, China.

Department of Mathematics, Simon Fraser University, Burnaby, BC, V5A 1S6, Canada.

出版信息

JMLR Workshop Conf Proc. 2013;28(2):37-45.

Abstract

Non-convex sparsity-inducing penalties have recently received considerable attentions in sparse learning. Recent theoretical investigations have demonstrated their superiority over the convex counterparts in several sparse learning settings. However, solving the non-convex optimization problems associated with non-convex penalties remains a big challenge. A commonly used approach is the Multi-Stage (MS) convex relaxation (or DC programming), which relaxes the original non-convex problem to a sequence of convex problems. This approach is usually not very practical for large-scale problems because its computational cost is a multiple of solving a single convex problem. In this paper, we propose a General Iterative Shrinkage and Thresholding (GIST) algorithm to solve the nonconvex optimization problem for a large class of non-convex penalties. The GIST algorithm iteratively solves a proximal operator problem, which in turn has a closed-form solution for many commonly used penalties. At each outer iteration of the algorithm, we use a line search initialized by the Barzilai-Borwein (BB) rule that allows finding an appropriate step size quickly. The paper also presents a detailed convergence analysis of the GIST algorithm. The efficiency of the proposed algorithm is demonstrated by extensive experiments on large-scale data sets.

摘要

非凸稀疏诱导惩罚最近在稀疏学习中受到了相当多的关注。最近的理论研究表明,在几种稀疏学习设置中,它们比凸惩罚具有优越性。然而,求解与非凸惩罚相关的非凸优化问题仍然是一个巨大的挑战。一种常用的方法是多阶段(MS)凸松弛(或DC规划),它将原始的非凸问题松弛为一系列凸问题。这种方法对于大规模问题通常不太实用,因为其计算成本是求解单个凸问题的倍数。在本文中,我们提出了一种通用迭代收缩阈值(GIST)算法,用于求解一大类非凸惩罚的非凸优化问题。GIST算法迭代地求解一个近端算子问题,而对于许多常用的惩罚,该近端算子问题又有一个闭式解。在算法的每次外层迭代中,我们使用由Barzilai-Borwein(BB)规则初始化的线搜索,该规则允许快速找到合适的步长。本文还给出了GIST算法的详细收敛性分析。通过在大规模数据集上的大量实验证明了所提算法的效率。

相似文献

2
Improving Sparsity and Scalability in Regularized Nonconvex Truncated-Loss Learning Problems.
IEEE Trans Neural Netw Learn Syst. 2018 Jul;29(7):2782-2793. doi: 10.1109/TNNLS.2017.2705429. Epub 2017 Jun 6.
3
The Proximal Alternating Minimization Algorithm for Two-Block Separable Convex Optimization Problems with Linear Constraints.
J Optim Theory Appl. 2019;182(1):110-132. doi: 10.1007/s10957-018-01454-y. Epub 2018 Dec 24.
4
Direct-Optimization-Based DC Dictionary Learning With the MCP Regularizer.
IEEE Trans Neural Netw Learn Syst. 2023 Jul;34(7):3568-3579. doi: 10.1109/TNNLS.2021.3114400. Epub 2023 Jul 6.
5
A Minibatch Proximal Stochastic Recursive Gradient Algorithm Using a Trust-Region-Like Scheme and Barzilai-Borwein Stepsizes.
IEEE Trans Neural Netw Learn Syst. 2021 Oct;32(10):4627-4638. doi: 10.1109/TNNLS.2020.3025383. Epub 2021 Oct 5.
7
Solving large-scale general phase retrieval problems via a sequence of convex relaxations.
J Opt Soc Am A Opt Image Sci Vis. 2018 Aug 1;35(8):1410-1419. doi: 10.1364/JOSAA.35.001410.
8
Nonconvex Nonsmooth Low Rank Minimization via Iteratively Reweighted Nuclear Norm.
IEEE Trans Image Process. 2016 Feb;25(2):829-39. doi: 10.1109/TIP.2015.2511584. Epub 2015 Dec 22.
9
Global Convergence Guarantees of (A)GIST for a Family of Nonconvex Sparse Learning Problems.
IEEE Trans Cybern. 2022 May;52(5):3276-3288. doi: 10.1109/TCYB.2020.3010960. Epub 2022 May 19.
10
Tighten after Relax: Minimax-Optimal Sparse PCA in Polynomial Time.
Adv Neural Inf Process Syst. 2014;2014:3383-3391.

引用本文的文献

1
Variable Selection for Multivariate Failure Time Data via Regularized Sparse-Input Neural Network.
Bioengineering (Basel). 2025 May 31;12(6):596. doi: 10.3390/bioengineering12060596.
2
Geometric deep learning for diffusion MRI signal reconstruction with continuous samplings (DISCUS).
Imaging Neurosci (Camb). 2024 Apr 2;2:1-18. doi: 10.1162/imag_a_00121. eCollection 2024 Apr 1.
3
Robust Support Vector Data Description with Truncated Loss Function for Outliers Depression.
Entropy (Basel). 2024 Jul 25;26(8):628. doi: 10.3390/e26080628.
4
Lp Quasi-norm Minimization: Algorithm and Applications.
Res Sq. 2023 Nov 28:rs.3.rs-3632062. doi: 10.21203/rs.3.rs-3632062/v1.
5
Integrative Learning of Structured High-Dimensional Data from Multiple Datasets.
Stat Anal Data Min. 2023 Apr;16(2):120-134. doi: 10.1002/sam.11601. Epub 2022 Nov 8.
6
Feature selection for kernel methods in systems biology.
NAR Genom Bioinform. 2022 Mar 7;4(1):lqac014. doi: 10.1093/nargab/lqac014. eCollection 2022 Mar.
7
Convex optimization algorithms in medical image reconstruction-in the age of AI.
Phys Med Biol. 2022 Mar 23;67(7). doi: 10.1088/1361-6560/ac3842.
8
Poisson noisy image restoration via overlapping group sparse and nonconvex second-order total variation priors.
PLoS One. 2021 Apr 20;16(4):e0250260. doi: 10.1371/journal.pone.0250260. eCollection 2021.
9
WHERE DID THE TUMOR START? AN INVERSE SOLVER WITH SPARSE LOCALIZATION FOR TUMOR GROWTH MODELS.
Inverse Probl. 2020 Apr;36(4). doi: 10.1088/1361-6420/ab649c. Epub 2020 Feb 26.
10
Neural Granger Causality.
IEEE Trans Pattern Anal Mach Intell. 2022 Aug;44(8):4267-4279. doi: 10.1109/TPAMI.2021.3065601. Epub 2022 Jul 1.

本文引用的文献

1
Multi-Stage Multi-Task Feature Learning.
Adv Neural Inf Process Syst. 2013 Oct;14:2979-3010.
2
Robust Multi-Task Feature Learning.
KDD. 2012 Aug 12;2012:895-903. doi: 10.1145/2339530.2339672.
3
Sparse Methods for Biomedical Data.
SIGKDD Explor. 2012 Jun 1;14(1):4-15. doi: 10.1145/2408736.2408739.
4
Robust face recognition via sparse representation.
IEEE Trans Pattern Anal Mach Intell. 2009 Feb;31(2):210-27. doi: 10.1109/TPAMI.2008.79.
5
Nonlinear image recovery with half-quadratic regularization.
IEEE Trans Image Process. 1995;4(7):932-46. doi: 10.1109/83.392335.
6
A simple and efficient algorithm for gene selection using sparse logistic regression.
Bioinformatics. 2003 Nov 22;19(17):2246-53. doi: 10.1093/bioinformatics/btg308.
7
The concave-convex procedure.
Neural Comput. 2003 Apr;15(4):915-36. doi: 10.1162/08997660360581958.

文献AI研究员

20分钟写一篇综述,助力文献阅读效率提升50倍。

立即体验

用中文搜PubMed

大模型驱动的PubMed中文搜索引擎

马上搜索

文档翻译

学术文献翻译模型,支持多种主流文档格式。

立即体验