A generic online acceleration scheme for optimization algorithms via relaxation and inertia

We propose generic acceleration schemes for a wide class of optimization and iterative schemes based on relaxation and inertia. In particular, we introduce methods that automatically tune the acceleration coefficients online and establish their convergence. This is made possible by considering class...

Full description

Saved in:
Bibliographic Details
Published in:Optimization methods & software 2019-03, Vol.34 (2), p.383-405
Main Authors: Iutzeler, F., Hendrickx, J. M.
Format: Article
Language:English
Subjects:
Acceleration
Algorithms
Applied optimization methods
Computer Science
Distributed, Parallel, and Cluster Computing
Engineering Sciences
Fixed points (mathematics)
Inertia
Machine Learning
Mathematics
Optimization
Optimization and Control
relaxation inertia
Signal and Image processing
Statistics
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:We propose generic acceleration schemes for a wide class of optimization and iterative schemes based on relaxation and inertia. In particular, we introduce methods that automatically tune the acceleration coefficients online and establish their convergence. This is made possible by considering classes of fixed-point iterations over averaged operators which encompass gradient methods, ADMM (Alternating Direction Method of Multipliers), primal dual algorithms and so on.
ISSN:1055-6788
1029-4937
DOI:10.1080/10556788.2017.1396601