A Theory-Driven Self-Labeling Refinement Method for Contrastive Representation Learning.
Pan ZhouCaiming XiongXiaotong YuanSteven Chu-Hong HoiPublished in: NeurIPS (2021)
Keyphrases
- learning algorithm
- detection method
- prior knowledge
- high accuracy
- preprocessing
- cost function
- learning scheme
- pairwise
- unsupervised learning
- significant improvement
- driven learning
- neural nets
- segmentation method
- active learning
- reinforcement learning
- similarity measure
- dynamic programming
- image representation
- clustering method
- learning process
- objective function
- learning tasks
- inference process
- machine learning