Adversarial training reduces safety of neural networks in robots: Research

Open in new window