Provable self-play algorithms for competitive reinforcement learning

Keyphrases

• Reinforcement Learning 100%
• Competitive Reinforcement 100%
• Self-play Algorithm 100%
• Regret 60%
• Self-play 40%
• Polynomial Time 20%
• Markov Decision Process 20%
• Value Iteration 20%
• Direct Supervision 20%
• Lower Confidence Bound 20%
• Sample Efficiency 20%
• Exploration-exploitation Trade-off 20%
• Markov Game 20%
• Superhuman Performance 20%

Computer Science

• Reinforcement Learning 100%
• Polynomial Time 20%
• Markov Decision Process 20%
• Confidence Bound 20%

Social Sciences

• Markov Decision Process 100%