A Symmetry-Preserving Extrapolated Primal-Dual Hybrid Gradient Method for Saddle-Point Problems
Axioms · 16.03.2026 · 15 (3) · 219
DOI: 10.3390/axioms15030219 ↗Kurz gefasst
Fachartikel in der Zeitschrift Axioms, online erschienen am 16.03.2026. Verfasst von Xiayang Zhang und 4 weiteren Autor:innen. Die Arbeit nennt 27 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
The primal-dual hybrid gradient (PDHG) method is widely used for convex–concave saddle-point problems, yet its extrapolated variants are typically asymmetric because only one side is extrapolated. We propose a symmetry-preserving refinement, E-PDHG, which performs dual-side extrapolation followed by an explicit correction step. Under standard step-size conditions, we establish global convergence for all η∈(−1,1) and derive a pointwise (non-ergodic) O(1/t) rate for the last iterate. The method does not improve the asymptotic complexity order of PDHG; instead, it enlarges the practically stable parameter region while retaining the same per-iteration cost. Numerical experiments on image deblurring/inpainting and additional machine learning benchmarks (logistic regression and LASSO) demonstrate improved finite-iteration stability and efficiency.
Zitieren
APA
Zhang, X., Li, W., Chang, B., Liu, W., & Zhang, S. (2026). A Symmetry-Preserving Extrapolated Primal-Dual Hybrid Gradient Method for Saddle-Point Problems. Axioms, 15(3), 219. https://doi.org/10.3390/axioms15030219
Vancouver
Zhang X, Li W, Chang B, Liu W, Zhang S. A Symmetry-Preserving Extrapolated Primal-Dual Hybrid Gradient Method for Saddle-Point Problems. Axioms. 2026;15(3):219. doi:10.3390/axioms15030219
BibTeX
@article{Zhang2026,
author = {Zhang, Xiayang and Li, Wenzhuo and Chang, Bowen and Liu, Wei and Zhang, Shiyu},
title = {A Symmetry-Preserving Extrapolated Primal-Dual Hybrid Gradient Method for Saddle-Point Problems},
journal = {Axioms},
year = {2026},
volume = {15},
number = {3},
pages = {219},
doi = {10.3390/axioms15030219}
}
EUVIMED