Universal-KD: Attention-based Output-Grounded Intermediate Layer Knowledge Distillation.
Yimeng WuMehdi RezagholizadehAbbas GhaddarMd. Akmal HaidarAli GhodsiPublished in: EMNLP (1) (2021)
Keyphrases
- domain knowledge
- knowledge transfer
- multi layer
- learning systems
- knowledge representation
- domain experts
- knowledge based systems
- prior knowledge
- expert systems
- knowledge base
- database
- case based reasoning
- knowledge acquisition
- reinforcement learning
- decision trees
- background knowledge
- expert knowledge
- information systems
- neural network