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
Это событие также входит в публичные календарные потоки ниже, поэтому посетители могут вернуться к более широкой серии, а не только к этой странице.
Время
12 Nov 2026 11:30 до 12 Nov 2026 12:30
Место
OAT, Room 11, ETH AI Center Foyer, Zurich
OAT, Room 11, ETH AI Center Foyer, Zurich, ch
Публикатор
ETH AI Center
Публичные календари с этим событием
Duration
1 час
Responsible
ETH AI Center
Source details
- Import status
- Regional Scan
- Source kind
- Внутренний
- Updated
- 2 часа назад
- Public calendars
- Switzerland Tech Events
Карта события
OAT, Room 11, ETH AI Center Foyer, Zurich, ch
Данные карты © участники 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