[Paper Reading] TabPFN
Автор: SMU AI Club
Загружено: 2026-02-25
Просмотров: 20
Описание:
Tabular data is very common among all disciplines. Traditional ML models aims to train algorithms by fitting a randomly initialized model onto a specific dataset. Newer work on TabPFNv2, TabICL, and MITRA not only shows pretrained model Prior-data Fitted neural Networks (PFNs) has the capability of predicting data what the model has never seen, they also perform better with in-context learning with row-attention. This seminar introduces what PFNs are, what can they do, and why they seem so magical.
TabPFN
TabPFNv2
TabPFN2.5
A Closer Look at TabPFNv2
MITRA
TabICL
Повторяем попытку...
Доступные форматы для скачивания:
Скачать видео
-
Информация по загрузке: