1 paper · 1 filter
Haomin Bai, Dingzhi Yu, Shuai Li +2
Group distributionally robust optimization (GDRO) aims to learn models that perform well across m distributions simultaneously. Existing methods either (i) query m samples per…