Alternating Proximal-Gradient Steps for (Stochastic) Nonconvex-Concave Minimax Problems
SIAM Journal on Optimization · 30.09.2023 · 33 (3) · 1884-1913
DOI: 10.1137/21m1465470 ↗Kurz gefasst
Fachartikel in der Zeitschrift SIAM Journal on Optimization, online erschienen am 02.08.2023. Verfasst von Radu Ioan Boţ und 1 weiteren Autor:innen. Die Arbeit wurde bisher 20-mal zitiert (Crossref) und nennt 55 Literaturangaben. Gefördert von: Austrian Science Fund.
Abstract (Originalsprache)
Minimax problems of the form [Formula: see text] have attracted increased interest largely due to advances in machine learning, in particular generative adversarial networks and adversarial learning. These are typically trained using variants of stochastic gradient descent for the two players. Although convex-concave problems are well understood with many efficient solution methods to choose from, theoretical guarantees outside of this setting are sometimes lacking even for the simplest algorithms. In particular, this is the case for alternating gradient descent ascent, where the two agents take turns updating their strategies. To partially close this gap in the literature we prove a novel global convergence rate for the stochastic version of this method for finding a critical point of [Formula: see text] in a setting which is not convex-concave.
Förderung
- Austrian Science Fund
Zitieren
APA
Boţ, R. I., & Böhm, A. (2023). Alternating Proximal-Gradient Steps for (Stochastic) Nonconvex-Concave Minimax Problems. SIAM Journal on Optimization, 33(3), 1884-1913. https://doi.org/10.1137/21m1465470
Vancouver
Boţ RI, Böhm A. Alternating Proximal-Gradient Steps for (Stochastic) Nonconvex-Concave Minimax Problems. SIAM Journal on Optimization. 2023;33(3):1884-1913. doi:10.1137/21m1465470
BibTeX
@article{Bo2023,
author = {Boţ, Radu Ioan and Böhm, Axel},
title = {Alternating Proximal-Gradient Steps for (Stochastic) Nonconvex-Concave Minimax Problems},
journal = {SIAM Journal on Optimization},
year = {2023},
volume = {33},
number = {3},
pages = {1884--1913},
doi = {10.1137/21m1465470}
}
EUVIMED