Leveraging Factored Action Spaces for Off-Policy Evaluation.
Aaman RebelloShengpu TangJenna WiensSonali ParbhooPublished in: CoRR (2023)
Keyphrases
- action space
- policy evaluation
- markov decision processes
- state space
- reinforcement learning
- policy iteration
- markov decision problems
- optimal policy
- temporal difference
- model free
- reinforcement learning algorithms
- finite state
- dynamic programming
- markov decision process
- least squares
- state variables
- partially observable markov decision processes
- monte carlo
- heuristic search
- markov chain
- state action
- action selection
- average reward
- real valued
- average cost
- variance reduction
- partially observable
- learning algorithm
- belief state
- reward function
- function approximation
- dynamical systems
- search space
- machine learning
- infinite horizon
- transition probabilities
- particle filter
- function approximators
- decision processes
- probability distribution
- probabilistic model
- bayesian networks