Robustness and Generalization via Generative Adversarial Training | IEEE Conference Publication | IEEE Xplore