Equilibrium stability drives cooperation in Q-learning algorithms

A new research paper explores how equilibrium stability can drive cooperation among Q-learners, particularly in scenarios where exploration does not vanish over time. The study focuses on the repeated Prisoner's Dilemma, analyzing the time-averaged fraction of cooperative strategies played by algorithms that continue to adapt. Researchers derived a boundary condition predicting when non-defection-dominated behavior emerges, which was validated through extensive simulations of epsilon-greedy Q-learning. AI

IMPACT Provides theoretical insights into cooperative strategies in adaptive reinforcement learning agents.

RANK_REASON Academic paper on reinforcement learning dynamics. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.MA (Multiagent) →

• epsilon-greedy Q-learning
• prisoner's dilemma
• Q-learning

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →