Mirror Descent on Riemannian Manifolds

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Jiang, Jiaxin, Shi, Lei, Tan, Jiyuan
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917351471972352
author Jiang, Jiaxin
Shi, Lei
Tan, Jiyuan
author_facet Jiang, Jiaxin
Shi, Lei
Tan, Jiyuan
contents Mirror Descent (MD) is a scalable first-order method widely used in large-scale optimization, with applications in image processing, policy optimization, and neural network training. This paper generalizes MD to optimization on Riemannian manifolds. In particular, we develop a Riemannian Mirror Descent (RMD) framework via reparameterization and further propose a stochastic variant of RMD. We also establish non-asymptotic convergence guarantees for both RMD and stochastic RMD. As an application to the Stiefel manifold, our RMD framework reduces to the Curvilinear Gradient Descent (CGD) method proposed in [26]. Moreover, when specializing the stochastic RMD framework to the Stiefel setting, we obtain a stochastic extension of CGD, which effectively addresses large-scale manifold optimization problems.
format Preprint
id arxiv_https___arxiv_org_abs_2603_17527
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Mirror Descent on Riemannian Manifolds
Jiang, Jiaxin
Shi, Lei
Tan, Jiyuan
Machine Learning
Optimization and Control
Mirror Descent (MD) is a scalable first-order method widely used in large-scale optimization, with applications in image processing, policy optimization, and neural network training. This paper generalizes MD to optimization on Riemannian manifolds. In particular, we develop a Riemannian Mirror Descent (RMD) framework via reparameterization and further propose a stochastic variant of RMD. We also establish non-asymptotic convergence guarantees for both RMD and stochastic RMD. As an application to the Stiefel manifold, our RMD framework reduces to the Curvilinear Gradient Descent (CGD) method proposed in [26]. Moreover, when specializing the stochastic RMD framework to the Stiefel setting, we obtain a stochastic extension of CGD, which effectively addresses large-scale manifold optimization problems.
title Mirror Descent on Riemannian Manifolds
topic Machine Learning
Optimization and Control
url https://arxiv.org/abs/2603.17527