Mirror Descent Methods with Weighting Scheme for Outputs for Constrained Variational Inequality Problems

Mohammad S. Alkousa, Belal A. Alashqar, Fedor S. Stonyakin, Tarek Nabhani, Seydamet S. Ablaev · 2025

Read Mirror Descent Methods with Weighting Scheme for Outputs for Constrained Variational Inequality Problems by Mohammad S. Alkousa, Belal A. Alashqar, Fedor S. Stonyakin, Tarek Nabhani, Seydamet S. Ablaev (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-2501-04034.txt.

Description

Mohammad Alkousa and co-authors speed up mirror descent methods for variational inequality problems, in a 2025 paper. Such problems crop up in games, traffic flow, and machine learning, wherever balance points are sought under rules. Mirror descent walks toward the answer in small steps shaped by a chosen geometry. How the steps' outputs are averaged can make or break the pace.

Their trick weights late points more than early ones, since late points sit closer to the prize. For problems with extra function-based limits, the method flips between bold and cautious steps as the limits allow. The authors prove the best pace known for this class, for a range of step rules. Bench tests on classic constrained cases beat a known projection method by a wide margin. The gains held across problem sizes.

Catalog record

CREATORMohammad S. Alkousa, Belal A. Alashqar, Fedor S. Stonyakin, Tarek Nabhani, Seydamet S. Ablaev
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE51 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2501-04034

Browse the full library · About the Tower of Babel