Variance-Reduced Fast Operator Splitting Methods for Generalized Equations

Quoc Tran-Dinh [email protected] · 2025

Read Variance-Reduced Fast Operator Splitting Methods for Generalized Equations by Quoc Tran-Dinh [email protected] (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-13046.txt.

Introduction

Generalized equations sit beneath minimization, minimax games, and inequality problems across optimization. Quoc Tran-Dinh develops in 2025 two fast splitting methods that also cut variance, the noise that plagues methods fed on random data samples. He first builds a family of variance-reduced estimators, covering famous schemes like SVRG, SAGA, and SARAH as special cases.

On top sits an accelerated forward-backward method that converges at the prized one-over-k-squared rate, and slightly better, with iterates settling on a true solution almost surely. Unusually, the methods tolerate a class of nonmonotone operators met in modern applications. A mirrored backward-forward variant matches the same guarantees and best-known costs.

Catalog record

CREATORQuoc Tran-Dinh [email protected]
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE212 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-13046

Browse the full library · About the Tower of Babel