Benign Overfitting without Linearity: Neural Network Classifiers Trained by Gradient Descent for Noisy Linear Data.

Spencer Frei, Niladri S. Chatterji, Peter L. Bartlett

Browse the full COLT paper archive.