Model-based RL in Contextual Decision Processes: PAC bounds and Exponential Improvements over Model-free Approaches

Keyphrases

• Model-free Methods 100%
• Decision Process 100%
• Sample Complexity 100%
• PAC Bounds 100%
• Model-based RL 100%
• Factored MDPs 66%
• Statistical Properties 33%
• Generic Model 33%
• Model Complexity 33%
• Sampling Efficiency 33%
• Model-based Reinforcement Learning 33%
• Bellman 33%
• Global Exploration 33%
• Exponential Separation 33%
• Strategic Exploration 33%
• Exponential Lower Bounds 33%
• Near-optimal Policy 33%
• Model-free Algorithms 33%
• Model-free RL 33%

Computer Science

• Decision Process 100%
• Statistical Property 50%
• Efficient Algorithm 50%
• Structural Parameter 50%
• Model-Based Reinforcement Learning 50%

Mathematics

• Decision Process 100%
• Statistical Property 50%
• Generality 50%
• Optimal Policy 50%
• Structural Parameter 50%