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
Tato událost je také součástí veřejných kalendářních proudů níže, takže se návštěvníci mohou vrátit k celé sérii.
Čas
12 Nov 2026 11:30 do 12 Nov 2026 12:30
Místo
OAT, Room 11, ETH AI Center Foyer, Zurich
OAT, Room 11, ETH AI Center Foyer, Zurich, ch
Vydavatel
ETH AI Center
Veřejné kalendáře s touto událostí
Duration
1 hodina
Responsible
ETH AI Center
Source details
- Import status
- Regional Scan
- Source kind
- Nativní
- Updated
- před 1 hodinou
- Public calendars
- Switzerland Tech Events
Mapa události
OAT, Room 11, ETH AI Center Foyer, Zurich, ch
Mapová data © přispěvatelé 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