One Policy is Enough: Parallel Exploration with a Single Policy is Near-Optimal for Reward-Free Reinforcement Learning

Keyphrases

• Near-optimal 100%
• Reinforcement Learning 100%
• Parallel Exploration 100%
• Parallelization 66%
• Markov Decision Process 66%
• Quantitative Effects 33%
• Minimax Optimal 33%
• Simple Procedure 33%
• Linear Speedup 33%
• Exploration Phase 33%
• Zero-sum Markov Game 33%

Computer Science

• Reinforcement Learning 100%
• Parallelism 66%
• Markov Decision Process 66%
• Linear Speedup 33%