Local convergence of alternating low‐rank optimization methods with overrelaxation

DOI10.1002/NLA.2459MaRDI QIDQ6133038zbMATH OpenOpenAlexWikidataFDO

Authors Ivan Oseledets, M. V. Rakhuba, André Uschmajew

Publication date 17 August 2023

Published in Numerical Linear Algebra with Applications (Search for Journal in Brave)

Copyright license Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International

Full work available at URL https://arxiv.org/abs/2111.14758

zbMATH Keywords

ALSoverrelaxationlow-rank optimizationSOR method

Mathematics Subject Classification ID

Iterative numerical methods for linear systems (65F10) Numerical methods for low-rank matrix approximation; matrix compression (65F55)

Abstract: The local convergence of alternating optimization methods with overrelaxation for low-rank matrix and tensor problems is established. The analysis is based on the linearization of the method which takes the form of an SOR iteration for a positive semidefinite Hessian and can be studied in the corresponding quotient geometry of equivalent low-rank representations. In the matrix case, the optimal relaxation parameter for accelerating the local convergence can be determined from the convergence rate of the standard method. This result relies on a version of Young's SOR theorem for positive semidefinite $2imes2$ block systems.

Recommendations

• Alternating least squares as moving subspace correction
• Local convergence of the alternating least squares algorithm for canonical tensor approximation
• Linear convergence of an alternating polar decomposition method for low rank orthogonal tensor approximations
• Provable accelerated gradient method for nonconvex low rank optimization
• On local convergence of alternating schemes for optimization of convex problems in the tensor train format

Cited in

(1)

• Local minima and convergence in low-rank semidefinite programming

This page was built for publication: Local convergence of alternating low‐rank optimization methods with overrelaxation

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6133038)