DORO: Distributional and Outlier Robust Optimization

            Runtian Zhai * 1 Chen Dan * 1 J. Zico Kolter 1 Pradeep Ravikumar 1

            Abstract                 shift problems, such as learning for algorithmic fairness
                                 (Dwork et al., 2012; Barocas & Selbst, 2016) where we
   Many machine learning tasks involve subpopu-        want to test model’s performance on key demographic sub-
   lation shift where the testing data distribution is     populations,  ...