arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2108.09093cs.LGcs.CV
收藏

对逆向鲁棒分类器生成能力的理解

Towards Understanding the Generative Capability of Adversarially Robust Classifiers

Zhejiang University(浙江大学)

Yao Zhu, Jiacheng Ma, Jiacheng Sun, Zewei Chen, Rongxin Jiang, Zhenguo Li

摘要:

最近,一些工作发现了一个有趣的现象,对抗鲁棒分类器可以生成与生成模型相当的好图像。我们从能量的角度研究这一现象,并提供了一个新的解释。我们重新制定了对抗性示例生成,对抗性训练和图像生成的能量函数。我们发现,对抗性训练有助于获得一个平坦的能量函数,并且在真实数据周围具有较低的能量,这是生成能力的关键。基于我们的新理解,我们进一步提出了一种更好的对抗训练方法,联合能量对抗训练(JEAT),它可以生成高质量的图像,并在广泛的攻击下实现新的最先进的鲁棒性。JEAT生成的图像的初始得分(CIFAR-10)为8.80,远远优于原始的鲁棒分类器(7.50)。

英文摘要:

Recently, some works found an interesting phenomenon that adversarially robust classifiers can generate good images comparable to generative models. We investigate this phenomenon from an energy perspective and provide a novel explanation. We reformulate adversarial example generation, adversarial training, and image generation in terms of an energy function. We find that adversarial training contributes to obtaining an energy function that is flat and has low energy around the real data, which is the key for generative capability. Based on our new understanding, we further propose a better adversarial training method, Joint Energy Adversarial Training (JEAT), which can generate high-quality images and achieve new state-of-the-art robustness under a wide range of attacks. The Inception Score of the images (CIFAR-10) generated by JEAT is 8.80, much better than original robust classifiers (7.50).

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑