Projected nonmonotone search methods for optimization with orthogonality constraints

被引：9

作者：

Dalmau Cedeno, Oscar Susano ^{[1
]}

Oviedo Leon, Harry Fernando ^{[1
]}

机构：

[1] CIMAT AC, Math Res Ctr, Guanajuato, Mexico

来源：

COMPUTATIONAL & APPLIED MATHEMATICS | 2018年 / 37卷 / 03期

关键词：

Constrained optimization; Orthogonality constraints; Non-monotone algorithm; Stiefel manifold; Optimization on manifolds; PRINCIPAL COMPONENT ANALYSIS; PROCRUSTES PROBLEM; CORRELATION-MATRICES; DIMENSION REDUCTION; STIEFEL MANIFOLD; RANK REDUCTION; MINIMIZATION; ALGORITHM; BARZILAI;

D O I：

10.1007/s40314-017-0501-6

中图分类号：

O29 [应用数学];

学科分类号：

070104 ;

摘要：

In this paper, we propose two feasible methods based on projections using a curvilinear search for solving optimization problems with orthogonality constraints. In one of them we apply a projected Adams-Moulton-like update scheme. All our algorithms compute the SVD decomposition in each iteration to preserve feasibility. Additionally, we present some convergence results. Finally, we perform numerical experiments with simulated problems; and analyze the performance of the proposed methods compared with state-of-the-art algorithms.

引用

页码：3118 / 3144

页数：27

共 50 条

[1] Projected nonmonotone search methods for optimization with orthogonality constraints
Oscar Susano Dalmau Cedeño
Harry Fernando Oviedo Leon
Computational and Applied Mathematics, 2018, 37 : 3118 - 3144
[2] NONMONOTONE CURVED SEARCH METHODS FOR UNCONSTRAINED OPTIMIZATION
邓乃扬
肖奕
周方俊
Numerical Mathematics A Journal of Chinese Universities(English Series), 1992, (01) : 34 - 48
[3] Lie group methods for optimization with orthogonality constraints
Plumbley, MD
INDEPENDENT COMPONENT ANALYSIS AND BLIND SIGNAL SEPARATION, 2004, 3195 : 1245 - 1252
[4] Nonmonotone inexact restoration approach for minimization with orthogonality constraints
Francisco, Juliano B.
Goncalves, Douglas S.
Bazan, Fermin S. V.
Paredes, Lila L. T.
NUMERICAL ALGORITHMS, 2021, 86 (04) : 1651 - 1684
[5] Nonmonotone inexact restoration approach for minimization with orthogonality constraints
Juliano B. Francisco
Douglas S. Gonçalves
Fermín S. V. Bazán
Lila L. T. Paredes
Numerical Algorithms, 2021, 86 : 1651 - 1684
[6] Nonmonotone curvilinear line search methods for unconstrained optimization
Univ of Wisconsin, Madison, United States
Comput Optim Appl, 2 (117-136):
[7] Quadratic Optimization with Orthogonality Constraints: Explicit Lojasiewicz Exponent and Linear Convergence of Line-Search Methods
Liu, Huikang
Wu, Weijie
So, Anthony Man-Cho
INTERNATIONAL CONFERENCE ON MACHINE LEARNING, VOL 48, 2016, 48
[8] PROJECTED NEWTON METHODS FOR OPTIMIZATION PROBLEMS WITH SIMPLE CONSTRAINTS
BERTSEKAS, DP
SIAM JOURNAL ON CONTROL AND OPTIMIZATION, 1982, 20 (02) : 221 - 246
[9] STRUCTURED QUASI-NEWTON METHODS FOR OPTIMIZATION WITH ORTHOGONALITY CONSTRAINTS
Hu, Jiang
Jiang, Bo
Lin, Lin
Wen, Zaiwen
Yuan, Ya-Xiang
SIAM JOURNAL ON SCIENTIFIC COMPUTING, 2019, 41 (04): : A2239 - A2269
[10] ACCELERATED OPTIMIZATION WITH ORTHOGONALITY CONSTRAINTS
Siegel, Jonathan W.
JOURNAL OF COMPUTATIONAL MATHEMATICS, 2021, 39 (02): : 207 - 226

← 1 2 3 4 5 →