Black-box Off-policy Estimation for Infinite-Horizon Reinforcement Learning.
Ali MousaviLihong LiQiang LiuDenny ZhouPublished in: ICLR (2020)
Keyphrases
- black box
- infinite horizon
- optimal policy
- reinforcement learning
- optimal control
- markov decision processes
- finite horizon
- partially observable
- markov decision process
- state space
- dynamic programming
- black boxes
- long run
- white box
- stochastic demand
- policy iteration
- production planning
- decision problems
- single item
- average cost
- function approximation
- integration testing
- learning algorithm
- dec pomdps
- test cases
- reinforcement learning algorithms
- model free
- finite state
- lost sales
- markov decision problems
- multistage
- sufficient conditions
- multi agent
- reward function
- action space
- inventory level
- machine learning
- markov chain
- white box testing