Are adversarial examples inevitable?

Ali Shafahi, W. Ronny Huang, Christoph Studer, Soheil Feizi, and Tom Goldstein, Are adversarial examples inevitable?, ICLR, 2019. (open review)

Theorem 1 (Existence of Adversarial Examples)
Theorem 2 (Adversarial examples on the cube)
Theorem 3 (Sparse adversarial examples)
Theorem 4 (Condition for existence of adversarial examples) 

For more on adversarial examples, try: (The Algorithm from MIT Tech Review, 2019/5/18)

• The full text of the new paper from MIT
• UC Berkeley professor Dawn Song’s talk at EmTech Digital and our coverage of it
• Researchers Ian Goodfellow, Jonathon Shlens, and Christian Szegedy’s foundational paper explaining adversarial examples
• OpenAI’s review of adversarial example research
• A step-by-step technical tutorial for how to create your own adversarial example
• And finally, my summary of acclaimed researcher Léon Bottou’s talk at ICLR, which teases how we might change the way we train deep-learning models to get rid of non-robust correlations