Can overfitted deep neural networks in adversarial training generalize? – An approximation viewpoint1 мар. 2024 г.
