Robust supervised learning with coordinate gradient descent

Stéphane Gaïffas, Ibrahim Merad

This paper considers the problem of supervised learning with linear methods when both features and labels can be corrupted, either in the form of heavy tailed data and/or corrupted…