Towards Understanding the Generative Capability of Adversarially Robust Classifiers
对抗性鲁棒分类器的生成能力研究
敵対的にロバストな分類器の生成能力の理解に向けて
적의 온건 분류 기의 생 성 능력 을 이해 하 다
Comprensión de la capacidad de generación de clasificadores robustos enemigos
Vers une compréhension de la capacité générative des classificateurs robustes à l'adversaire
понимание способности противника к созданию стабилизатора
Yao Zhu ¹, Jiacheng Ma ², Jiacheng Sun ², Zewei Chen ², Rongxin Jiang 蒋荣欣 ¹, Zhenguo Li ²
¹ Zhejiang University
浙江大学
² Huawei Noah’s Ark Lab
华为诺亚方舟实验室
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). In particular, we achieve new state-of-the-art robustness on CIFAR-10 (from 57.20% to 62.04%) and CIFAR-100 (from 30.03% to 30.18%) without extra training data.
