Nearly Minimax Optimal Reinforcement Learning for Discounted MDPs.
Jiafan HeDongruo ZhouQuanquan GuPublished in: NeurIPS (2021)
Keyphrases
- markov decision processes
- reinforcement learning
- optimal policy
- dynamic programming
- average reward
- average cost
- finite horizon
- markov decision process
- state space
- optimal control
- action sets
- infinite horizon
- total reward
- worst case
- finite state
- discounted reward
- reinforcement learning algorithms
- approximate dynamic programming
- policy iteration
- function approximation
- stationary policies
- discount factor
- reward function
- state and action spaces
- factored mdps
- control policies
- semi markov decision processes
- partially observable
- optimal solution
- decision theoretic planning
- continuous state spaces
- model based reinforcement learning
- reinforcement learning methods
- long run
- decision problems
- policy evaluation
- policy search
- markov decision problems
- model free
- optimal strategy
- markov chain
- linear programming
- multi agent