Adversarial Examples Unveiled
Florian sheds light on the mysterious nature of adversarial examples, revealing that they might be inherent to the data itself. By separating robust and non-robust features, researchers demonstrate how classifiers can be trained to be resilient to adversarial attacks based on the underlying data characteristics.In this clip
From this podcast

Machine Learning Street Talk (MLST)
#040 - Adversarial Examples (Dr. Nicholas Carlini, Dr. Wieland Brendel, Florian Tramèr)
Related Questions