Off-Policy Evaluation for Large Action Spaces via Embeddings.
Yuta SaitoThorsten JoachimsPublished in: ICML (2022)
Keyphrases
- policy evaluation
- action space
- markov decision processes
- reinforcement learning
- policy iteration
- state space
- temporal difference
- markov decision problems
- action selection
- optimal policy
- function approximation
- model free
- markov decision process
- finite state
- dynamic programming
- reinforcement learning algorithms
- least squares
- state action
- monte carlo
- average cost
- function approximators
- partially observable
- np hard
- multi agent
- learning algorithm
- single agent
- reward function
- semi parametric
- machine learning
- average reward
- real valued