Nearly Minimax Optimal Regret for Learning Infinite-horizon Average-reward MDPs with Linear Function Approximation.
Yue WuDongruo ZhouQuanquan GuPublished in: AISTATS (2022)
Keyphrases
- reinforcement learning
- average reward
- total reward
- function approximation
- markov decision processes
- optimal policy
- infinite horizon
- actor critic
- stochastic games
- policy iteration
- td learning
- partially observable
- long run
- finite horizon
- optimal control
- policy gradient
- function approximators
- temporal difference
- average cost
- state action
- model free
- state space
- discount factor
- reinforcement learning algorithms
- dynamic programming
- discounted reward
- learning tasks
- learning algorithm
- action selection
- markov decision process
- evaluation function
- machine learning
- finite state
- rl algorithms
- machine learning algorithms
- policy evaluation
- transfer learning
- decision theoretic
- reinforcement learning methods
- markov decision problems
- linear programming
- decision problems
- action space
- search algorithm
- radial basis function
- partially observable markov decision processes