Semi-Supervised Offline Reinforcement Learning with Action-Free Trajectories.
Qinqing ZhengMikael HenaffBrandon AmosAditya GroverPublished in: CoRR (2022)
Keyphrases
- semi supervised
- reinforcement learning
- action selection
- view invariant
- supervised learning
- action space
- semi supervised learning
- reward shaping
- multi view
- partially observable domains
- labeled data
- state action
- function approximation
- state space
- learning algorithm
- unlabeled data
- unsupervised learning
- transition model
- trajectory data
- moving objects
- semi supervised classification
- active learning
- co training
- model free
- pairwise
- fitted q iteration
- machine learning
- multi agent
- sensory inputs
- reinforcement learning algorithms
- learning process
- human actions
- transfer learning
- semi supervised clustering
- spatio temporal
- markov decision process
- pairwise constraints
- continuous state
- policy search
- optimal policy