Unifying Gradient Estimators for Meta-Reinforcement Learning via Off-Policy Evaluation.
Yunhao TangTadashi KozunoMark RowlandRémi MunosMichal ValkoPublished in: NeurIPS (2021)
Keyphrases
- policy evaluation
- gradient estimators
- reinforcement learning
- temporal difference
- least squares
- model free
- function approximation
- monte carlo
- markov decision processes
- td learning
- policy iteration
- variance reduction
- reinforcement learning algorithms
- optimal policy
- semi parametric
- machine learning
- state space
- dynamic programming
- learning algorithm
- evaluation function
- partially observable markov decision processes
- multi agent
- statistical inference
- image processing
- markov decision problems
- finite state
- edge detector
- supervised learning
- computer vision
- neural network