Graphon Mean-Field Subsampling for Cooperative Heterogeneous Multi-Agent Reinforcement Learning

Emile Anand, Richard Hoffmann, Sarah Liaw, Adam Wierman · 2026

Read Graphon Mean-Field Subsampling for Cooperative Heterogeneous Multi-Agent Reinforcement Learning by Emile Anand, Richard Hoffmann, Sarah Liaw, Adam Wierman (2026) 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-2602-16196.txt.

Description

Coordinating large crowds of interacting agents is hard because the joint state-action space grows exponentially with agent count. Emile Anand and coauthors introduce GMFS, a graphon mean-field subsampling design for scalable cooperative multi-agent reinforcement learning with mixed interactions. Mean-field methods ease the burden by aggregating interactions, but they assume uniform interactions. Graphon designs capture mixed interactions yet grow costly as agent numbers rise.

GMFS subsamples agents by interaction strength, approximating the graphon-weighted mean field. It learns a policy with sample complexity polynomial in the subsample size and a gap to optimal that shrinks with the square root of that size. Numerical simulations in robotic coordination verify the theory, with GMFS achieving near-optimal results. The design thus offers a scalable route to cooperative learning where agents differ and crowds are large.

Catalog record

CREATOREmile Anand, Richard Hoffmann, Sarah Liaw, Adam Wierman
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE168 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602-16196

Browse the full library · About the Tower of Babel