Exploiting Action Impact Regularity and Exogenous State Variables for Offline Reinforcement Learning.
Vincent LiuJames R. WrightMartha WhitePublished in: J. Artif. Intell. Res. (2023)
Keyphrases
- state variables
- state space
- reinforcement learning
- action space
- dynamic systems
- action selection
- partially observable domains
- reward shaping
- dynamic programming
- markov decision processes
- reinforcement learning algorithms
- random variables
- markov decision process
- optimal policy
- partially observable
- state action
- transition model
- factored markov decision processes
- function approximation
- particle filter
- learning algorithm
- reward function
- initial state
- planning problems
- heuristic search
- temporal difference
- search algorithm
- multi agent
- fitted q iteration
- machine learning