Year
Month
(Preprint) Towards Understanding the Generative Capability of Adversarially Robust Classifiers
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
arXiv, 2021-08-20
Abstract

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.
Towards Understanding the Generative Capability of Adversarially Robust Classifiers_1
Towards Understanding the Generative Capability of Adversarially Robust Classifiers_2
Towards Understanding the Generative Capability of Adversarially Robust Classifiers_3
  • Dynamic interactive bitwise meta-holography with ultra-high computational and display frame rates
  • Yuncheng Liu, Ke Xu, Xuhao Fan, Xinger Wang, Xuan Yu, Wei Xiong, Hui Gao
  • Opto-Electronic Advances
  • 2024-01-25
  • Multi-dimensional multiplexing optical secret sharing framework with cascaded liquid crystal holograms
  • Keyao Li, Yiming Wang, Dapu Pi, Baoli Li, Haitao Luan, Xinyuan Fang, Peng Chen, Yanqing Lu, Min Gu
  • Opto-Electronic Advances
  • 2024-01-25
  • Physics-informed deep learning for fringe pattern analysis
  • Wei Yin, Yuxuan Che, Xinsheng Li, Mingyu Li, Yan Hu, Shijie Feng, Edmund Y. Lam, Qian Chen, Chao Zuo
  • Opto-Electronic Advances
  • 2024-01-25
  • Advancing computer-generated holographic display thanks to diffraction model-driven deep nets
  • Vittorio Bianco, Pietro Ferraro
  • Opto-Electronic Advances
  • 2024-01-16
  • Inverse design for material anisotropy and its application for a compact X-cut TFLN on-chip wavelength demultiplexer
  • Jiangbo Lyu, Tao Zhu, Yan Zhou, Zhenmin Chen, Yazhi Pi, Zhengtong Liu, Xiaochuan Xu, Ke Xu, Xu Ma, Lei Wang, Zizheng Cao, Shaohua Yu
  • Opto-Electronic Science
  • 2024-01-09
  • Improved spatiotemporal resolution of anti-scattering super-resolution label-free microscopy via synthetic wave 3D metalens imaging
  • Yuting Xiao, Lianwei Chen, Mingbo Pu, Mingfeng Xu, Qi Zhang, Yinghui Guo, Tianqu Chen, Xiangang Luo
  • Opto-Electronic Science
  • 2024-01-05
  • Wide-spectrum optical synthetic aperture imaging via spatial intensity interferometry
  • Chunyan Chu, Zhentao Liu, Mingliang Chen, Xuehui Shao, Guohai Situ, Yuejin Zhao, Shensheng Han
  • Opto-Electronic Advances
  • 2023-3-10
  • Flat soliton microcomb source
  • Xinyu Wang, Xuke Qiu, Mulong Liu, Feng Liu, Mengmeng Li, Linpei Xue, Bohan Chen, Mingran Zhang, Peng Xie
  • Opto-Electronic Science
  • 2023-12-29
  • Smart palm-size optofluidic hematology analyzer for automated imaging-based leukocyte concentration detection
  • Deer Su, Xiangyu Li, Weida Gao, Qiuhua Wei, Haoyu Li, Changliang Guo, Weisong Zhao
  • Opto-Electronic Science
  • 2023-12-28
  • Exceptional-point-enhanced sensing in an all-fiber bending sensor
  • Zheng Li, Jingxu Chen, Lingzhi Li, Jiejun Zhang, Jianping Yao
  • Opto-Electronic Advances
  • 2023-12-12
  • All-optical object identification and three-dimensional reconstruction based on optical computing metasurface
  • Dingyu Xu, Wenhao Xu, Qiang Yang, Wenshuai Zhang, Shuangchun Wen, Hailu Luo,
  • Opto-Electronic Advances
  • 2023-12-12



  • Context-Aware Candidates for Image Cropping                                Image Cropping Assisted By Modeling Inter-Patch Relations
    About
    |
    Contact
    |
    Copyright © PubCard