Lecture
This lecture covers the theory and applications of adversarial machine learning, focusing on minmax optimization, adversarial training, generative adversarial networks, and the challenges of robustness to adversarial examples. The instructor discusses the formulation of adversarial examples, the difficulty of minmax optimization, and the use of different norms in adversarial attacks. The lecture also explores the robustness of classifiers in high-dimensional spaces, the impact of adversarial examples on neural networks, and the practical implementation of adversarial training. Various optimization techniques, such as primal-dual optimization and stochastic subgradient descent, are presented in the context of adversarial machine learning.