SURF: Semi-supervised Reward Learning with Data Augmentation for Feedback-efficient Preference-based Reinforcement Learning.
Jongjin ParkYounggyo SeoJinwoo ShinHonglak LeePieter AbbeelKimin LeePublished in: ICLR (2022)
Keyphrases
- reinforcement learning
- semi supervised
- data sets
- labeled data
- learning process
- prior knowledge
- unlabeled data
- supervised learning
- database
- learning problems
- data collection
- learning algorithm
- co training
- data analysis
- semi supervised learning
- eligibility traces
- background knowledge
- partially observable environments
- state space
- unsupervised learning
- multi view
- function approximation
- reinforcement learning methods
- data sources
- training data