Adversarial Examples Explained
Andrew discusses the intriguing phenomenon of adversarial examples in vision, where minor perturbations to images can lead to significant misclassifications by machine learning models. He highlights the conventional belief that these examples are merely bugs, while emphasizing the importance of distinguishing between useful and useless features learned by neural networks. Through a simple experiment, insights into the vulnerabilities of these models are explored, challenging existing notions of their robustness.In this clip
From this podcast
Machine Learning Street Talk (MLST)
Adversarial Examples and Data Modelling - Andrew Ilyas (MIT)
Related Questions