Last iterate convergence in no-regret learning: constrained min-max optimization for convex-concave landscapes

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Lei, Qi, Nagarajan, Sai Ganesh, Panageas, Ioannis, Wang, Xiao
Natura: Preprint
Pubblicazione: 2020
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866914059218059264
author Lei, Qi
Nagarajan, Sai Ganesh
Panageas, Ioannis
Wang, Xiao
author_facet Lei, Qi
Nagarajan, Sai Ganesh
Panageas, Ioannis
Wang, Xiao
contents In a recent series of papers it has been established that variants of Gradient Descent/Ascent and Mirror Descent exhibit last iterate convergence in convex-concave zero-sum games. Specifically, \cite{DISZ17, LiangS18} show last iterate convergence of the so called "Optimistic Gradient Descent/Ascent" for the case of \textit{unconstrained} min-max optimization. Moreover, in \cite{Metal} the authors show that Mirror Descent with an extra gradient step displays last iterate convergence for convex-concave problems (both constrained and unconstrained), though their algorithm does not follow the online learning framework; it uses extra information rather than \textit{only} the history to compute the next iteration. In this work, we show that "Optimistic Multiplicative-Weights Update (OMWU)" which follows the no-regret online learning framework, exhibits last iterate convergence locally for convex-concave games, generalizing the results of \cite{DP19} where last iterate convergence of OMWU was shown only for the \textit{bilinear case}. We complement our results with experiments that indicate fast convergence of the method.
format Preprint
id arxiv_https___arxiv_org_abs_2002_06768
institution arXiv
publishDate 2020
record_format arxiv
spellingShingle Last iterate convergence in no-regret learning: constrained min-max optimization for convex-concave landscapes
Lei, Qi
Nagarajan, Sai Ganesh
Panageas, Ioannis
Wang, Xiao
Machine Learning
Computer Science and Game Theory
In a recent series of papers it has been established that variants of Gradient Descent/Ascent and Mirror Descent exhibit last iterate convergence in convex-concave zero-sum games. Specifically, \cite{DISZ17, LiangS18} show last iterate convergence of the so called "Optimistic Gradient Descent/Ascent" for the case of \textit{unconstrained} min-max optimization. Moreover, in \cite{Metal} the authors show that Mirror Descent with an extra gradient step displays last iterate convergence for convex-concave problems (both constrained and unconstrained), though their algorithm does not follow the online learning framework; it uses extra information rather than \textit{only} the history to compute the next iteration. In this work, we show that "Optimistic Multiplicative-Weights Update (OMWU)" which follows the no-regret online learning framework, exhibits last iterate convergence locally for convex-concave games, generalizing the results of \cite{DP19} where last iterate convergence of OMWU was shown only for the \textit{bilinear case}. We complement our results with experiments that indicate fast convergence of the method.
title Last iterate convergence in no-regret learning: constrained min-max optimization for convex-concave landscapes
topic Machine Learning
Computer Science and Game Theory
url https://arxiv.org/abs/2002.06768