Minimax Optimal Reinforcement Learning for Discounted MDPs.
Jiafan HeDongruo ZhouQuanquan GuPublished in: CoRR (2020)
Keyphrases
- markov decision processes
- reinforcement learning
- dynamic programming
- optimal policy
- average cost
- average reward
- finite horizon
- markov decision process
- state space
- optimal control
- total reward
- infinite horizon
- action sets
- discounted reward
- finite state
- stationary policies
- policy iteration
- state and action spaces
- action space
- reinforcement learning algorithms
- worst case
- function approximation
- model based reinforcement learning
- discount factor
- approximate dynamic programming
- continuous state spaces
- multi agent
- model free
- factored mdps
- decision theoretic planning
- reward function
- control policy
- temporal difference
- optimal solution
- action selection
- policy evaluation
- control policies
- partially observable markov decision processes
- markov decision problems
- evaluation function
- decision problems
- reinforcement learning methods
- state dependent