Abstract: Recent research has shown that deep learning models are likely to make incorrect predictions even when exposed to minor perturbations. To address this, training models on adversarial ...