The sketched landing method for large-scale optimization under orthogonality constraints

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Goyens, Florentin, Mataigne, Simon, Absil, Pierre-Antoine
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911733462859776
author Goyens, Florentin
Mataigne, Simon
Absil, Pierre-Antoine
author_facet Goyens, Florentin
Mataigne, Simon
Absil, Pierre-Antoine
contents We propose the \emph{sketched landing method}, a randomized variant of the landing method for optimization under orthogonality constraints. Each landing step consists of the sum of a \emph{normal} component, which reduces infeasibility, and a \emph{tangent} component, which decreases the objective function. Our main contribution is the introduction of low-dimensional random \emph{sketch matrices} to reduce the computational cost of these directions. We consider both dense (Gaussian) and sparse (subsampling) sketch matrices, and show how they reduce the per-iteration cost while preserving convergence guarantees in expectation.
format Preprint
id arxiv_https___arxiv_org_abs_2605_31505
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle The sketched landing method for large-scale optimization under orthogonality constraints
Goyens, Florentin
Mataigne, Simon
Absil, Pierre-Antoine
Optimization and Control
49M29 (primary), 53C21, 65K10, 90C30 (secondary)
We propose the \emph{sketched landing method}, a randomized variant of the landing method for optimization under orthogonality constraints. Each landing step consists of the sum of a \emph{normal} component, which reduces infeasibility, and a \emph{tangent} component, which decreases the objective function. Our main contribution is the introduction of low-dimensional random \emph{sketch matrices} to reduce the computational cost of these directions. We consider both dense (Gaussian) and sparse (subsampling) sketch matrices, and show how they reduce the per-iteration cost while preserving convergence guarantees in expectation.
title The sketched landing method for large-scale optimization under orthogonality constraints
topic Optimization and Control
49M29 (primary), 53C21, 65K10, 90C30 (secondary)
url https://arxiv.org/abs/2605.31505