Conceptadversarial robustness
Adversarial Robustness
An important research direction in deep learning security aimed at improving the resistance of neural networks to adversarial example attacks.
Timeline (last 90 days)
Oct 25
If a decision function has a small Lipschitz constant, small perturbations of the input will not cause drastic changes in the output; this corresponds to the smoothness of the decision surface and is one of the core concepts in adversarial robustness research.
Unverified50%