An Improved Last-Iterate Convergence Rate for Anchored Gradient Descent Ascent

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Surina, Anja, Suggala, Arun, Tsoukalas, George, Kovsharov, Anton, Shirobokov, Sergey, Ruiz, Francisco J. R., Kohli, Pushmeet, Chaudhuri, Swarat
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914446027259904
author Surina, Anja
Suggala, Arun
Tsoukalas, George
Kovsharov, Anton
Shirobokov, Sergey
Ruiz, Francisco J. R.
Kohli, Pushmeet
Chaudhuri, Swarat
author_facet Surina, Anja
Suggala, Arun
Tsoukalas, George
Kovsharov, Anton
Shirobokov, Sergey
Ruiz, Francisco J. R.
Kohli, Pushmeet
Chaudhuri, Swarat
contents We analyze the last-iterate convergence of the Anchored Gradient Descent Ascent algorithm for smooth convex-concave min-max problems. While previous work established a last-iterate rate of $\mathcal{O}(1/t^{2-2p})$ for the squared gradient norm, where $p \in (1/2, 1)$, it remained an open problem whether the improved exact $\mathcal{O}(1/t)$ rate is achievable. In this work, we resolve this question in the affirmative. This result was discovered autonomously by an AI system capable of writing formal proofs in Lean. The Lean proof can be accessed at https://github.com/google-deepmind/formal-conjectures/pull/3675/commits/a13226b49fd3b897f4c409194f3bcbeb96a08515
format Preprint
id arxiv_https___arxiv_org_abs_2604_03782
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle An Improved Last-Iterate Convergence Rate for Anchored Gradient Descent Ascent
Surina, Anja
Suggala, Arun
Tsoukalas, George
Kovsharov, Anton
Shirobokov, Sergey
Ruiz, Francisco J. R.
Kohli, Pushmeet
Chaudhuri, Swarat
Optimization and Control
Artificial Intelligence
We analyze the last-iterate convergence of the Anchored Gradient Descent Ascent algorithm for smooth convex-concave min-max problems. While previous work established a last-iterate rate of $\mathcal{O}(1/t^{2-2p})$ for the squared gradient norm, where $p \in (1/2, 1)$, it remained an open problem whether the improved exact $\mathcal{O}(1/t)$ rate is achievable. In this work, we resolve this question in the affirmative. This result was discovered autonomously by an AI system capable of writing formal proofs in Lean. The Lean proof can be accessed at https://github.com/google-deepmind/formal-conjectures/pull/3675/commits/a13226b49fd3b897f4c409194f3bcbeb96a08515
title An Improved Last-Iterate Convergence Rate for Anchored Gradient Descent Ascent
topic Optimization and Control
Artificial Intelligence
url https://arxiv.org/abs/2604.03782