Fast Rates for the Regret of Offline Reinforcement Learning.
Yichun HuNathan KallusMasatoshi UeharaPublished in: COLT (2021)
Keyphrases
- reinforcement learning
- total reward
- reward function
- function approximation
- lower bound
- learning process
- markov decision processes
- model free
- reinforcement learning algorithms
- loss function
- multi agent reinforcement learning
- multi armed bandit
- learning algorithm
- state space
- transfer learning
- weighted majority
- online learning
- real time
- learning problems
- action selection
- temporal difference
- optimal policy
- robotic control
- reinforcement learning methods
- expert advice
- minimax regret
- confidence bounds
- transition model
- average reward
- pairwise
- machine learning
- data sets