Sample-Efficient Reinforcement Learning for Linearly-Parameterized MDPs with a Generative Model

Keyphrases

• Markov Decision Process 100%
• Generative Models 100%
• Model-based Approach 100%
• Sample-Efficient Reinforcement Learning 100%
• Optimal Policy 66%
• Q-learning 66%
• High Probability 33%
• Reinforcement Learning 33%
• Discount Factor 33%
• State Space 33%
• Tight 33%
• Sample Complexity 33%
• Nearly Optimal 33%
• Complexity Scale 33%
• Complexity Bounds 33%
• Sampling Efficiency 33%
• State Action 33%
• Curse of Dimensionality 33%
• Minimax Optimal 33%
• Action Space 33%
• Feature Dimension 33%
• Minimax Lower Bound 33%
• Transition Kernel 33%
• Q-function 33%
• Probability Transition 33%
• Action Features 33%
• Optimal Sample Complexity 33%

Computer Science

• Reinforcement Learning 100%
• Markov Decision Process 100%
• Generative Model 100%
• State Space 33%
• Feature Dimension 33%
• Match Approach 33%

Mathematics

• Coordinate Basis 33%