Interpretation of Neural Networks is Susceptible to Universal Adversarial Perturbations

Open in new window