MASTER: Multi-task Pre-trained Bottlenecked Masked Autoencoders Are Better Dense Retrievers.
Kun ZhouXiao LiuYeyun GongWayne Xin ZhaoDaxin JiangNan DuanJi-Rong WenPublished in: ECML/PKDD (2) (2023)
Keyphrases
- multi task
- pre trained
- multi task learning
- training data
- learning tasks
- denoising
- feature selection
- training examples
- multi class
- learning problems
- transfer learning
- sparse learning
- control signals
- labeled data
- text categorization
- information gain
- data mining
- text classification
- supervised learning
- computer vision
- machine learning