5-Recent-Developments-in-Robustness-against-Adversarial-Perturbation

This project sums up the main points of the recent developments in deep learning for robustness against adversarial perturbation, then elaborates the tasks, datasets, evaluations of the previous approach adversarial training and the state of the art parseval regularization. Furthermore, a detailed comparison was made between of the previous methods and the most advanced approaches. A summary is provided at the end of this work, aiming to provide more new ideas and directions for future work.