Global-Local Regularization Via Distributional Robustness

Hoang Phan, Trung Le, Trung Phung +3

Despite superior performance in many situations, deep neural networks are often vulnerable to adversarial examples and distribution shifts, limiting model generalization ability in…