Hybrid domain adaptation method defends deep learning models against multi-targeted adversarial examples
Researchers report a hybrid adversarial domain adaptation method combining logit correlation and gradient randomization that improved accuracy by 2 to ...
