AI Paper Atlas

Research map / Adversarial robustness and security / Attacks on text and graphs

Towards Understanding the Generative Capability of Adversarially Robust Classifiers

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

ICCV 2021 · Adversarial robustness and security › Attacks on text and graphs · approach: Defenses against universal perturbations · 14 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add adversarial training to a normal classifier (lower joint and marginal Fisher scores)WideResNet-28-10 classifier · CIFAR-10Joint Fisher Score: 6.97 -> 2.83Improved
use E(x,y) only instead of E(x,y)-E(x) when generating images and training adversarial examples (PreJEAT) (ablation)adversarially robust classifier · CIFAR-10 generationimage quality (qualitative)Improved
add joint likelihood training of p(x,y) with SGLD to PreJEAT (JEAT vs adversarial training)robust classifier · CIFAR-10 generationInception Score: 7.50 -> 8.80Improved
JEAT instead of adversarial trainingrobust classifier · CIFAR-10 generationFID: 60.90 -> 38.24Improved
JEAT instead of JEMhybrid generative classifier · CIFAR-10 generationInception Score: 8.66 -> 8.80Improved
JEAT instead of JEMhybrid generative classifier · CIFAR-10 generationFID: 38.40 -> 38.24No clear effect
JEAT instead of JEMhybrid generative classifier · CIFAR-10 PGD-20 eps=8/255robust accuracy (%): 6.11% -> 30.55%Improved
JEAT instead of JEMhybrid generative classifier · CIFAR-10standard accuracy (%): 92.90% -> 85.16%Hurt
JEAT instead of JEMhybrid generative classifier · CIFAR-10Joint Fisher Score: 0.330 -> 0.023Improved
JEAT instead of JEMhybrid generative classifier · CIFAR-10Marginal Fisher Score: 1.44 -> 0.15Improved
attack using only E(x,y) instead of full cross-entropy loss (ablation)normal classifier WRN-28-10 · CIFAR-10accuracy under attack (%): 0.12% -> 0.11%No clear effect
inject stronger Gaussian noise during normal classifier training (ablation)normal classifier · CIFAR-10 generationimage recognizability (qualitative): noise std 8/255 -> 24/255Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy