Publications · Journal article · J. Nonlinear Var. Anal. 2026
An accelerated primal–dual algorithm with backtracking for decentralized constrained optimization
Journal of Nonlinear and Variational Analysis (special issue in honor of Yurii Nesterov), 10(3), pp. 555–596, 2026.
In brief
A primal–dual method for constrained problems whose data are spread over a network of agents, each holding its own objective and constraints and communicating only with neighbors. The method is accelerated, and a backtracking rule lets it find usable stepsizes on the fly instead of requiring the problem’s Lipschitz constants in advance, with convergence rates for the resulting decentralized iteration.
Cite
@article{xu2026backtracking,
title = {An accelerated primal–dual algorithm with backtracking for decentralized constrained optimization},
author = {Qiushui Xu and Necdet Serhat Aybat and Mert Gürbüzbalaban},
year = {2026},
journal = {Journal of Nonlinear and Variational Analysis (special issue in honor of Yurii Nesterov)},
volume = {10(3)},
pages = {555--596},
url = {https://jnva.biemdas.com/issues/JNVA2026-3-4.pdf},
}Relatedsame topics