Outlier Robust Adversarial Training

Shu Hu, Zhenhuan Yang, Xin Wang +2

Supervised learning models are challenged by the intrinsic complexities of training data such as outliers and minority subpopulations and intentional attacks at inference time with…