Conservative Q-Learning for Offline Reinforcement Learning.
Aviral KumarAurick ZhouGeorge TuckerSergey LevinePublished in: NeurIPS (2020)
Keyphrases
- reinforcement learning
- function approximation
- reinforcement learning algorithms
- state space
- model free
- state action space
- multi agent
- stochastic approximation
- temporal difference learning
- learning algorithm
- temporal difference
- action selection
- optimal policy
- markov decision processes
- real time
- dynamic programming
- control problems
- optimal control
- learning capabilities
- partially observable
- state action
- reinforcement learning methods
- actor critic
- continuous state and action spaces
- neural network
- learning agent
- machine learning
- continuous state
- multi agent reinforcement learning
- state abstraction
- hierarchical reinforcement learning
- eligibility traces
- supervised learning