Research map / Adversarial robustness and security / Certified and empirical robustness
Explaining Adversarial Robustness of Neural Networks from Clustering Effect Perspective
Yulin Jin, Xiaoyu Zhang, Jian Lou, Xu Ma, Zilong Wang, Xiaofeng Chen
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add SAT generalized inner maximization and clustering regularizer to PGD-AT | PGD-AT ResNet-18 · CIFAR10, LAFEAT100 | robust accuracy (%): 37.05 -> 43.08 | Improved |
| combine SAT with TRADES | TRADES ResNet-18 · CIFAR10, LAFEAT100 | robust accuracy (%): 42.02 -> 45.35 | Improved |
| combine SAT with AWP-AT | AWP-AT WRN28x10 · CIFAR10, LAFEAT100 | robust accuracy (%): 48.61 -> 54.70 | Improved |
| combine SAT with AWP-AT | AWP-AT ResNet-18 · CIFAR10, PGD100 | robust accuracy (%): 47.07 -> 48.82 | Improved |
| SAT alone vs AWP-AT and TRADES on output-layer attacks | SAT ResNet-18 · CIFAR10, PGD100 | robust accuracy (%): SAT 42.92 vs TRADES 44.88 | No clear effect |
| replace TRADES with SAT-TRADES with more generated data | TRADES · CIFAR10 500K generated, AutoAttack | robust accuracy (%): 62.82 -> 62.51 | No clear effect |
| run attacks to convergence (1000 iterations) | SAT · CIFAR10 ResNet-18 | robust accuracy (%): LAFEAT100 43.08 -> LAFEAT1000 42.49 | No clear effect |
| add clustering regularization L_Cluster | SAT training · SVHN ResNet-18 | convergence | No clear effect |
Closest papers by meaning
- Clustering Effect of Adversarial Robust Models (NeurIPS 2021 · similarity 0.91)
- Improving Adversarial Robustness by Enforcing Local and Global Compactness (ECCV 2020 · similarity 0.90)
- Adversarial Learning for Robust Deep Clustering (NeurIPS 2020 · similarity 0.89)
- On the Robustness of Deep Clustering Models: Adversarial Attacks and Defenses (NeurIPS 2022 · similarity 0.88)
- Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data (ICLR 2025 · similarity 0.88)
- On the Adversarial Robustness of Multi-Kernel Clustering (ICML 2025 · similarity 0.87)
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.
