Safety-Guaranteed, Accelerated Learning in MDPs with Local Side Information

Keyphrases

• Transition Probability 100%
• Optimal Control Strategy 100%
• Accelerated Learning 100%
• On-board Sensors 100%
• Numerical Experiments 50%
• Learning Process 50%
• Physical Systems 50%
• Learning Algorithm 50%
• Environment Modeling 50%
• Real-world Problems 50%
• Learning Improvement 50%
• Current Estimation 50%
• Value of Information 50%
• Uncertain Dynamics 50%
• Markov Decision Process 50%
• Dynamic Synthesis 50%
• Expected Improvement 50%
• Learning Speed 50%
• Classical Learning 50%
• Mars Rover 50%
• Indirect Sampling 50%
• Unknown Transition Probabilities 50%
• Policy Mandate 50%
• Optimal Control Synthesis 50%

Computer Science

• Side Information 100%
• Accelerated Learning 100%
• Learning Process 20%
• Physical System 20%
• Environment Model 20%
• Real-World Problem 20%
• Learning Algorithm 20%
• Information Value 20%
• Markov Decision Process 20%
• Policy Mandate 20%