Feb. 13, 2024, 5:41 a.m. | Yiwei Ding Alexander Lerch

cs.LG updates on arXiv.org arxiv.org

Common knowledge distillation methods require the teacher model and the student model to be trained on the same task. However, the usage of embeddings as teachers has also been proposed for different source tasks and target tasks. Prior work that uses embeddings as teachers ignores the fact that the teacher embeddings are likely to contain irrelevant knowledge for the target task. To address this problem, we propose to use an embedding compression module with a trainable teacher transformation to obtain …

compression cs.lg distillation embedding embeddings knowledge prior tasks teachers transfer usage work

Research Scholar (Technical Research)

@ Centre for the Governance of AI | Hybrid; Oxford, UK

HPC Engineer (x/f/m) - DACH

@ Meshcapade GmbH | Remote, Germany

Data Engineering Director-Big Data technologies (Hadoop, Spark, Hive, Kafka)

@ Visa | Bengaluru, India

Senior Data Engineer

@ Manulife | Makati City, Manulife Philippines Head Office

GDS Consulting Senior Data Scientist 2

@ EY | Taguig, PH, 1634

IT Data Analyst Team Lead

@ Rosecrance | Rockford, Illinois, United States