Variance Reduced Random Relaxed Projection Method for Constrained Finite-Sum Minimization Problems

被引:2
|
作者
Yang, Zhichun [1 ]
Xia, Fu-quan [1 ]
Tu, Kai [2 ]
Yue, Man-Chung [3 ,4 ]
机构
[1] Sichuan Normal Univ, Sch Math Sci, Chengdu 610068, Peoples R China
[2] Shenzhen Univ, Sch Math Sci, Shenzhen 518060, Peoples R China
[3] Univ Hong Kong, Musketeers Fdn Inst Data Sci, Hong Kong, Peoples R China
[4] Univ Hong Kong, Dept Ind & Mfg Syst Engn, Hong Kong, Peoples R China
基金
中国国家自然科学基金;
关键词
Signal processing algorithms; Convergence; Approximation algorithms; Optimization; Convex functions; Indexes; Standards; Constrained optimization; finite-sum minimization; random projection method; relaxed projection; variance reduction; convergence rate analysis; ERROR-BOUNDS; OPTIMIZATION; CONVERGENCE; DESCENT;
D O I
10.1109/TSP.2024.3386388
中图分类号
TM [电工技术]; TN [电子技术、通信技术];
学科分类号
0808 ; 0809 ;
摘要
For many applications in signal processing and machine learning, we are tasked with minimizing a large sum of convex functions subject to a large number of convex constraints. In this paper, we devise a new random projection method (RPM) to efficiently solve this problem. Compared with existing RPMs, our proposed algorithm features two useful algorithmic ideas. First, at each iteration, instead of projecting onto the set defined by one of the constraints, our algorithm only requires projecting onto a half-space approximation of the set, which significantly reduces the computational cost as it admits a closed-form formula. Second, to exploit the structure that the objective is a sum, variance reduction is incorporated into our algorithm to further improve the performance. As theoretical contributions, under a novel error bound condition and other standard assumptions, we prove that the proposed RPM converges to an optimal solution and that both optimality and feasibility gaps vanish at a sublinear rate. In particular, via a new analysis framework, we show that our RPM attains a faster convergence rate in optimality gap than existing RPMs when the objective function has a Lipschitz continuous gradient, capitalizing the benefit of the variance reduction. We also provide sufficient conditions for the error bound condition to hold. Experiments on a beamforming problem and a robust classification problem are presented to demonstrate the superiority of our RPM over existing ones.
引用
收藏
页码:2188 / 2203
页数:16
相关论文
共 50 条
  • [41] Method with batching for stochastic finite-sum variational inequalities in non-Euclidean setting
    Pichugin, Alexander
    Pechin, Maksim
    Beznosikov, Aleksandr
    Novitskii, Vasilii
    Gasnikov, Alexander
    CHAOS SOLITONS & FRACTALS, 2024, 187
  • [42] A relaxed projection method for solving bilevel variational inequality problems
    Anh, Pham Ngoc
    Khanh, Phan Quoc
    Truong, Nguyen Duc
    OPTIMIZATION, 2024,
  • [43] Residual selection in a projection method for convex minimization problems
    Cegielski, A
    Dylewski, R
    OPTIMIZATION, 2003, 52 (02) : 211 - 220
  • [44] An interactive satisficing method through the variance minimization model for fuzzy random multiobjective linear programming problems
    Katagiri, H
    Sakawa, M
    Ohsaki, S
    MULTI-OBJECTIVE PROGRAMMING AND GOAL PROGRAMMING, 2003, : 171 - 176
  • [45] An Iterative Method for Equilibrium and Constrained Convex Minimization Problems
    Yazdi, Maryam
    Shabani, Mohammad Mehdi
    Sababe, Saeed Hashemi
    KYUNGPOOK MATHEMATICAL JOURNAL, 2022, 62 (01): : 89 - 106
  • [46] Constrained maximum variance stopping for a finite horizon increasing random walk
    Fontem, Belleh
    OPERATIONS RESEARCH LETTERS, 2020, 48 (04) : 493 - 499
  • [47] Regularized gradient-projection methods for equilibrium and constrained convex minimization problems
    Ming Tian
    Li-Hua Huang
    Journal of Inequalities and Applications, 2013
  • [48] Regularized gradient-projection methods for equilibrium and constrained convex minimization problems
    Tian, Ming
    Huang, Li-Hua
    JOURNAL OF INEQUALITIES AND APPLICATIONS, 2013,
  • [49] The Stable Finite Element Method for Minimization Problems
    Liu, B.
    Huang, Y.
    JOURNAL OF COMPUTATIONAL AND THEORETICAL NANOSCIENCE, 2008, 5 (07) : 1251 - 1254
  • [50] A Variation on a Random Coordinate Minimization Method for Constrained Polynomial Optimization
    Calafiore, Giuseppe C.
    Possieri, Corrado
    IEEE CONTROL SYSTEMS LETTERS, 2018, 2 (03): : 531 - 536