Provably Robust Deep Learning via Adversarially Trained Smoothed Classifiers

Provably Robust Deep Learning via Adversarially Trained Smoothed Classifiers

Explore the concept of provably robust deep learning through adversarially trained smoothed classifiers in this 47-minute lecture by Jerry Li from Microsoft Research. Delve into key topics including randomization, the Zico idea, experimental results, semi-supervised learning, training techniques, notation, gradients, and the optimal gradient. Examine the full algorithm, its parameters, and the resulting outcomes. Gain insights into the frontiers of deep learning and the development of more resilient neural networks.