no code implementations • 13 Sep 2022 • Thomas Cordier, Victor Bouvier, Gilles Hénaff, Céline Hudelot
Machine Learning models are prone to fail when test data are different from training data, a situation often encountered in real applications known as distribution shift.