Compositional Conservatism: A Transductive Approach in Offline Reinforcement Learning.
Yeda SongDongwook LeeGunhee KimPublished in: ICLR (2024)
Keyphrases
- reinforcement learning
- function approximation
- text classification
- state space
- real time
- semi supervised
- multi agent
- support vector
- machine learning
- transductive learning
- reinforcement learning algorithms
- mobile robot
- markov decision processes
- model free
- temporal difference
- optimal policy
- temporal difference learning
- robotic control
- learning algorithm
- unlabeled data
- learning problems
- supervised learning
- optimal control
- dynamic programming
- action selection
- active learning
- training data
- learning capabilities
- markov decision process
- information systems
- reinforcement learning methods
- multi agent reinforcement learning
- policy search
- data mining