Why Adam work in optimizing language models: heavy-tailed data imbalance
Adam and related methods are the default optimizers in training language models because gradient descent is so slow it is thought it simply does not work. Many methods have been proposed to improve on Adam, but we have a poor understanding of why it works better in the first plac
Cet événement fait aussi partie des flux publics de calendrier ci-dessous, afin que les visiteurs reviennent à la série complète.
Horaire
12 Nov 2026 11:30 à 12 Nov 2026 12:30
Lieu
OAT, Room 11, ETH AI Center Foyer, Zurich
OAT, Room 11, ETH AI Center Foyer, Zurich, ch
Diffuseur
ETH AI Center
Calendriers publics contenant cet événement
Duration
1 heure
Responsible
ETH AI Center
Source details
- Import status
- Regional Scan
- Source kind
- Natif
- Updated
- il y a 1 heure
- Public calendars
- Switzerland Tech Events
Carte de l’événement
OAT, Room 11, ETH AI Center Foyer, Zurich, ch
Données cartographiques © contributeurs OpenStreetMap
Adam and related methods are the default optimizers in training language models because gradient descent is so slow it is thought it simply does not work. Many methods have been proposed to improve on Adam, but we have a poor understanding of why it works better in the first place. This talk will be about understanding what makes language models hard for gradient descent, and how Adam fixes it. We Price / conditions: free