论文标题
猫：定制的对抗训练，以改善鲁棒性
CAT: Customized Adversarial Training for Improved Robustness
论文作者
论文摘要
对抗性训练已成为改善神经网络鲁棒性的最有效方法之一。但是，它通常遭受清洁和干扰数据的概括不佳。在本文中，我们提出了一种新算法，名为“定制对抗训练”（CAT），该算法可自适应地自定义扰动水平和对抗培训中每个培训样本的相应标签。我们表明，通过广泛的实验，所提出的算法比以前的对抗训练方法可以实现更好和鲁棒的精度。
Adversarial training has become one of the most effective methods for improving robustness of neural networks. However, it often suffers from poor generalization on both clean and perturbed data. In this paper, we propose a new algorithm, named Customized Adversarial Training (CAT), which adaptively customizes the perturbation level and the corresponding label for each training sample in adversarial training. We show that the proposed algorithm achieves better clean and robust accuracy than previous adversarial training methods through extensive experiments.
