Textbook of AI

Abstract. Discovers adversarial examples: small, often imperceptible perturbations to images that cause deep networks to produce wildly incorrect outputs. The paper opened the field of adversarial robustness research.