Network Analysis on the Symmetric Coordination in a Reinforcement-Learning-Based Minority Game

Area of Science:

• Econophysics
• Complex Systems
• Network Science

Background:

• The Minority Game (MG) is a key model for studying inductive reasoning and self-organization in multi-agent systems.
• Traditional MG coordination involves symmetry breaking and role polarization among agents.
• Recent research introduced policy-based reinforcement learning for novel symmetric coordination without role specialization.

Purpose of the Study:

• To thoroughly analyze the novel symmetric coordination emerging from reinforcement learning in the Minority Game.
• To apply complex network tools for a deep understanding of this emergent coordination.
• To investigate the structural, functional, and temporal characteristics of this coordination pattern.

Main Methods:

• Construction of correlation networks among agents in the Minority Game.
• Application of complex network analysis techniques (structural, functional, temporal).
• Analysis of emergent coordination patterns in a multi-agent system with reinforcement learning.

Main Results:

• Confirmation of preserved symmetry at the collective level in the Minority Game.
• Identification of a consistent and robust form of distributed coordination.
• Demonstration of network-based approaches for understanding emergent order.

Conclusions:

• Reinforcement learning enables symmetric coordination in the Minority Game without role segregation.
• Complex network analysis effectively reveals the nature of emergent distributed coordination.
• This study highlights the power of network science in analyzing adaptive multi-agent systems.