VAIN: Attentional Multi-agent Predictive Modeling

TL;DR

VAIN is a new attentional model for multi-agent prediction that scales linearly with agents and outperforms existing methods in chess and soccer domains.

Contribution

Introduces VAIN, an attentional architecture for multi-agent modeling that improves scalability and effectiveness over previous interaction-based models.

Findings

VAIN outperforms existing multi-agent models in chess and soccer.

The model scales linearly with the number of agents.

Effective for complex multi-agent physical and social systems.

Abstract

Multi-agent predictive modeling is an essential step for understanding physical, social and team-play systems. Recently, Interaction Networks (INs) were proposed for the task of modeling multi-agent physical systems, INs scale with the number of interactions in the system (typically quadratic or higher order in the number of agents). In this paper we introduce VAIN, a novel attentional architecture for multi-agent predictive modeling that scales linearly with the number of agents. We show that VAIN is effective for multi-agent predictive modeling. Our method is evaluated on tasks from challenging multi-agent prediction domains: chess and soccer, and outperforms competing multi-agent approaches.