Metalearning-Based Alternating Minimization Algorithm for Nonconvex Optimization

被引:37
|
作者
Xia, Jing-Yuan [1 ]
Li, Shengxi [2 ]
Huang, Jun-Jie [1 ]
Yang, Zhixiong [1 ]
Jaimoukha, Imad M. [3 ]
Gunduz, Deniz [3 ,4 ]
机构
[1] Natl Univ Def Technol, Coll Elect Engn, Coll Comp, Changsha 410073, Peoples R China
[2] Beihang Univ, Coll Elect Engn, Beijing 100191, Peoples R China
[3] Imperial Coll London, Dept Elect & Elect Engn, London SW7 2AZ, England
[4] Univ Modena & Reggio Emilia UNIMORE, Dept Engn Enzo Ferrari, I-41121 Modena, Italy
基金
中国国家自然科学基金;
关键词
Optimization; Minimization; Task analysis; Deep learning; Neural networks; Signal processing algorithms; Iterative algorithms; Alternating minimization (AM); deep unfolding; gaussian mixture model (GMM); matrix completion; meta-learning (ML); MAXIMUM-LIKELIHOOD; MATRIX COMPLETION; DECONVOLUTION; MODEL;
D O I
10.1109/TNNLS.2022.3165627
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
In this article, we propose a novel solution for nonconvex problems of multiple variables, especially for those typically solved by an alternating minimization (AM) strategy that splits the original optimization problem into a set of subproblems corresponding to each variable and then iteratively optimizes each subproblem using a fixed updating rule. However, due to the intrinsic nonconvexity of the original optimization problem, the optimization can be trapped into a spurious local minimum even when each subproblem can be optimally solved at each iteration. Meanwhile, learning-based approaches, such as deep unfolding algorithms, have gained popularity for nonconvex optimization; however, they are highly limited by the availability of labeled data and insufficient explainability. To tackle these issues, we propose a meta-learning based alternating minimization (MLAM) method that aims to minimize a part of the global losses over iterations instead of carrying minimization on each subproblem, and it tends to learn an adaptive strategy to replace the handcrafted counterpart resulting in advance on superior performance. The proposed MLAM maintains the original algorithmic principle, providing certain interpretability. We evaluate the proposed method on two representative problems, namely, bilinear inverse problem: matrix completion and nonlinear problem: Gaussian mixture models. The experimental results validate the proposed approach outperforms AM-based methods.
引用
收藏
页码:5366 / 5380
页数:15
相关论文
共 50 条
  • [1] Explainable Attention Pruning: A Metalearning-Based Approach
    Rajapaksha P.
    Crespi N.
    IEEE Transactions on Artificial Intelligence, 2024, 5 (06): : 2505 - 2516
  • [2] MetaMP: Metalearning-Based Multipatch Image Aesthetics Assessment
    Yang, Jiachen
    Zhou, Yanshuang
    Zhao, Yang
    Lu, Wen
    Gao, Xinbo
    IEEE TRANSACTIONS ON CYBERNETICS, 2023, 53 (09) : 5716 - 5728
  • [3] A preconditioned alternating minimization framework for nonconvex and half quadratic optimization
    Deng, Shengxiang
    Ayed, Ismail Ben
    Sun, Hongpeng
    arXiv, 2021,
  • [4] A Metalearning-based Sparse Aperture ISAR Imaging Method
    Xia J.
    Yang Z.
    Zhou Z.
    Liao H.
    Zhang S.
    Fu Y.
    Journal of Radars, 2023, 12 (03) : 849 - 859
  • [5] Convergent Nested Alternating Minimization Algorithms for Nonconvex Optimization Problems
    Gur, Eyal
    Sabach, Shoham
    Shtern, Shimrit
    MATHEMATICS OF OPERATIONS RESEARCH, 2023, 48 (01) : 53 - 77
  • [6] A Learned Proximal Alternating Minimization Algorithm and Its Induced Network for a Class of Two-Block Nonconvex and Nonsmooth Optimization
    Chen, Yunmei
    Liu, Lezhi
    Zhang, Lei
    JOURNAL OF SCIENTIFIC COMPUTING, 2025, 103 (02)
  • [7] A Stochastic Proximal Alternating Minimization for Nonsmooth and Nonconvex
    Driggs, Derek
    Tang, Junqi
    Liang, Jingwei
    Davies, Mike
    Schonlieb, Carola-Bibiane
    SIAM JOURNAL ON IMAGING SCIENCES, 2021, 14 (04): : 1932 - 1970
  • [8] ON ITERATED MINIMIZATION IN NONCONVEX OPTIMIZATION
    JONGEN, HT
    MOBERT, T
    TAMMER, K
    MATHEMATICS OF OPERATIONS RESEARCH, 1986, 11 (04) : 679 - 691
  • [9] Two-step inertial Bregman alternating minimization algorithm for nonconvex and nonsmooth problems
    Jing Zhao
    Qiao-Li Dong
    Michael Th. Rassias
    Fenghui Wang
    Journal of Global Optimization, 2022, 84 : 941 - 966
  • [10] Two-step inertial Bregman alternating minimization algorithm for nonconvex and nonsmooth problems
    Zhao, Jing
    Dong, Qiao-Li
    Rassias, Michael Th
    Wang, Fenghui
    JOURNAL OF GLOBAL OPTIMIZATION, 2022, 84 (04) : 941 - 966