Quantum Adversarial Attack

I gave a talk at the 2021 American Physical Society (APS) March Meeting on our paper Robust in Practice: Adversarial Attacks on Quantum Machine Learning.

Below are extended slides showing that we argue against the extreme, exponential vulnerabilities of general quantum classifiers theoretically, and instead prove mild, polynomial vulnerabilities on practically relevant states to be classified, suggesting that general quantum classifiers are still potentially useful.