5 ms·Re-weighted gradient descent via distributionally robust optimization1 points by mfiguiere 3y ago