Non-Markovian Reward Modelling from Trajectory Labels via Interpretable Multiple Instance Learning.
Joseph EarlyTom BewleyChristine EversSarvapali D. RamchurnPublished in: NeurIPS (2022)
Keyphrases
- multiple instance learning
- reinforcement learning
- multiple instance
- image categorization
- class labels
- positive bags
- supervised learning
- object based image retrieval
- multi class
- image annotation
- reward function
- semi supervised
- diverse density
- pairwise
- semi supervised learning
- multi label
- instance selection
- training examples
- training data
- machine learning