论文标题

通过对抗性攻击$ \ Mathcal {H}δ\ Mathcal {H} $ - Divergence通过对抗性攻击进行多步域的适应

Multi-step domain adaptation by adversarial attack to $\mathcal{H} Δ\mathcal{H}$-divergence

论文作者

Asadulaev, Arip, Panfilov, Alexander, Filchenkov, Andrey

论文摘要

对抗示例在不同模型之间可以转移。在我们的论文中,我们建议将此属性用于多步域适应。在无监督的域适应设置中,我们证明,用对抗性示例代替源域$ \ Mathcal {h}Δ\ Mathcal {h} $ - divergence可以提高目标域的源分类器精度。我们的方法可以连接到大多数域适应技术。我们进行了一系列实验,并在数字和办公室的数据集上的准确性提高了。

Adversarial examples are transferable between different models. In our paper, we propose to use this property for multi-step domain adaptation. In unsupervised domain adaptation settings, we demonstrate that replacing the source domain with adversarial examples to $\mathcal{H} Δ\mathcal{H}$-divergence can improve source classifier accuracy on the target domain. Our method can be connected to most domain adaptation techniques. We conducted a range of experiments and achieved improvement in accuracy on Digits and Office-Home datasets.

扫码加入交流群

加入微信交流群

微信交流群二维码

扫码加入学术交流群,获取更多资源