all AI news
Heavy-Tailed Class Imbalance and Why Adam Outperforms Gradient Descent on Language Models
March 1, 2024, 5:43 a.m. | Frederik Kunstner, Robin Yadav, Alan Milligan, Mark Schmidt, Alberto Bietti
cs.LG updates on arXiv.org arxiv.org
Abstract: Adam has been shown to outperform gradient descent in optimizing large language transformers empirically, and by a larger margin than on other tasks, but it is unclear why this happens. We show that the heavy-tailed class imbalance found in language modeling tasks leads to difficulties in the optimization dynamics. When training with gradient descent, the loss associated with infrequent words decreases slower than the loss associated with frequent ones. As most samples come from relatively …
abstract adam arxiv class cs.cl cs.lg found gradient language language models large language leads math.oc modeling show stat.ml tasks transformers type
More from arxiv.org / cs.LG updates on arXiv.org
Jobs in AI, ML, Big Data
Data Architect
@ University of Texas at Austin | Austin, TX
Data ETL Engineer
@ University of Texas at Austin | Austin, TX
Lead GNSS Data Scientist
@ Lurra Systems | Melbourne
Senior Machine Learning Engineer (MLOps)
@ Promaton | Remote, Europe
Tableau/PowerBI Developer (A.Con)
@ KPMG India | Bengaluru, Karnataka, India
Software Engineer, Backend - Data Platform (Big Data Infra)
@ Benchling | San Francisco, CA