Certifying Concavity and Monotonicity in Games via Sum-of-Squares Hierarchies

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Leon, Vincent, Sakos, Iosif, Sim, Ryann, Varvitsiotis, Antonios
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908704962510848
author Leon, Vincent
Sakos, Iosif
Sim, Ryann
Varvitsiotis, Antonios
author_facet Leon, Vincent
Sakos, Iosif
Sim, Ryann
Varvitsiotis, Antonios
contents Concavity and its refinements underpin tractability in multiplayer games, where players independently choose actions to maximize their own payoffs which depend on other players' actions. In concave games, where players' strategy sets are compact and convex, and their payoffs are concave in their own actions, strong guarantees follow: Nash equilibria always exist and decentralized algorithms converge to equilibria. If the game is furthermore monotone, an even stronger guarantee holds: Nash equilibria are unique under strictness assumptions. Unfortunately, we show that certifying concavity or monotonicity is NP-hard, already for games where utilities are multivariate polynomials and compact, convex basic semialgebraic strategy sets -- an expressive class that captures extensive-form games with imperfect recall. On the positive side, we develop two hierarchies of sum-of-squares programs that certify concavity and monotonicity of a given game, and each level of the hierarchies can be solved in polynomial time. We show that almost all concave/monotone games are certified at some finite level of the hierarchies. Subsequently, we introduce SOS-concave/monotone games, which globally approximate concave/monotone games, and show that for any given game we can compute the closest SOS-concave/monotone game in polynomial time. Finally, we apply our techniques to canonical examples of imperfect recall extensive-form games.
format Preprint
id arxiv_https___arxiv_org_abs_2512_10292
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Certifying Concavity and Monotonicity in Games via Sum-of-Squares Hierarchies
Leon, Vincent
Sakos, Iosif
Sim, Ryann
Varvitsiotis, Antonios
Computer Science and Game Theory
Multiagent Systems
Optimization and Control
Concavity and its refinements underpin tractability in multiplayer games, where players independently choose actions to maximize their own payoffs which depend on other players' actions. In concave games, where players' strategy sets are compact and convex, and their payoffs are concave in their own actions, strong guarantees follow: Nash equilibria always exist and decentralized algorithms converge to equilibria. If the game is furthermore monotone, an even stronger guarantee holds: Nash equilibria are unique under strictness assumptions. Unfortunately, we show that certifying concavity or monotonicity is NP-hard, already for games where utilities are multivariate polynomials and compact, convex basic semialgebraic strategy sets -- an expressive class that captures extensive-form games with imperfect recall. On the positive side, we develop two hierarchies of sum-of-squares programs that certify concavity and monotonicity of a given game, and each level of the hierarchies can be solved in polynomial time. We show that almost all concave/monotone games are certified at some finite level of the hierarchies. Subsequently, we introduce SOS-concave/monotone games, which globally approximate concave/monotone games, and show that for any given game we can compute the closest SOS-concave/monotone game in polynomial time. Finally, we apply our techniques to canonical examples of imperfect recall extensive-form games.
title Certifying Concavity and Monotonicity in Games via Sum-of-Squares Hierarchies
topic Computer Science and Game Theory
Multiagent Systems
Optimization and Control
url https://arxiv.org/abs/2512.10292