Implicit steepest descent algorithm for optimization with orthogonality constraints

Optimization problems with orthogonality constraints appear widely in applications from science and engineering. We address these types of problems from a numerical approach. Our new framework combines the steepest gradient descent, using implicit information, with a projection operator in order to...

Full description

Saved in:

Bibliographic Details
Published in:	Optimization letters Vol. 16; no. 6; pp. 1773 - 1797
Main Author:	Oviedo, Harry
Format:	Journal Article
Language:	English
Published:	Berlin/Heidelberg Springer Berlin Heidelberg 01.07.2022
Subjects:	Computational Intelligence Mathematics Mathematics and Statistics Numerical and Computational Physics Operations Research/Decision Theory Optimization Original Paper Simulation Eigenvalue problem Orthogonality constrained optimization Total energy minimization Gradient type methods Riemannian optimization
ISSN:	1862-4472, 1862-4480
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!

Description
Summary:	Optimization problems with orthogonality constraints appear widely in applications from science and engineering. We address these types of problems from a numerical approach. Our new framework combines the steepest gradient descent, using implicit information, with a projection operator in order to construct a feasible sequence of points. In addition, we adopt an adaptive Barzilai–Borwein steplength mixed with a globalization technique in order to speed-up the convergence of our procedure. The global convergence, and some theoretical related to our algorithm are proved. The effectiveness of our proposed algorithm is demonstrated on a variety of problems including Rayleigh quotient maximization, heterogeneous quadratics minimization, weighted orthogonal procrustes problems and total energy minimization. Numerical results show that the new procedure can outperform some state of the art solvers on some practically problems.
ISSN:	1862-4472 1862-4480
DOI:	10.1007/s11590-021-01801-5