Abstract
The Stochastic Primal-Dual Hybrid Gradient (SPDHG) was proposed by Chambolle et al. (2018) and is an efficient algorithm to solve some nonsmooth large-scale optimization problems. In this paper we prove its almost sure convergence for convex but not necessarily strongly convex functionals. We also look into its application to parallel Magnetic Resonance Imaging reconstruction in order to test performance of SPDHG. Our numerical results show that for a range of settings SPDHG converges significantly faster than its deterministic counterpart.
Original language | English |
---|---|
Title of host publication | Scale Space and Variational Methods in Computer Vision - 8th International Conference, SSVM 2021, Proceedings |
Editors | Abderrahim Elmoataz, Jalal Fadili, Yvain Quéau, Julien Rabin, Loïc Simon |
Place of Publication | Germany |
Publisher | Springer Science and Business Media Deutschland GmbH |
Pages | 254-266 |
Number of pages | 13 |
ISBN (Print) | 9783030755485 |
DOIs | |
Publication status | Published - 29 Apr 2021 |
Event | 8th International Conference on Scale Space and Variational Methods in Computer Vision, SSVM 2021 - Virtual, Online Duration: 16 May 2021 → 20 May 2021 |
Publication series
Name | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) |
---|---|
Volume | 12679 LNCS |
ISSN (Print) | 0302-9743 |
ISSN (Electronic) | 1611-3349 |
Conference
Conference | 8th International Conference on Scale Space and Variational Methods in Computer Vision, SSVM 2021 |
---|---|
City | Virtual, Online |
Period | 16/05/21 → 20/05/21 |
Funding
MJE and CD acknowledge support from the EPSRC (EP/S026045/1). MJE is also supported by EPSRC (EP/T026693/1), the Faraday Institution (EP/T007745/1) and the Leverhulme Trust (ECF-2019-478). EBG acknowledges the Mexican Council of Science and Technology (CONACyT).
Keywords
- Convex optimization
- Inverse problems
- Parallel magnetic resonance imaging
- Primal-dual algorithm
- Stochastic optimization
ASJC Scopus subject areas
- Theoretical Computer Science
- General Computer Science