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
Este evento también forma parte de los flujos públicos de calendario de abajo, para que los visitantes vuelvan a la serie completa y no solo a esta página.
Horario
12 Nov 2026 11:30 a 12 Nov 2026 12:30
Lugar
OAT, Room 11, ETH AI Center Foyer, Zurich
OAT, Room 11, ETH AI Center Foyer, Zurich, ch
Publicador
ETH AI Center
Calendarios públicos que incluyen este evento
Duration
1 hora
Responsible
ETH AI Center
Source details
- Import status
- Regional Scan
- Source kind
- Nativo
- Updated
- hace 1 hora
- Public calendars
- Switzerland Tech Events
Mapa del evento
OAT, Room 11, ETH AI Center Foyer, Zurich, ch
Datos del mapa © colaboradores de 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