Research map / Adversarial robustness and security / Certified and empirical robustness
Adversarial Training is a Form of Data-dependent Operator Norm Regularization
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| use data-dependent spectral norm regularization instead of standard training | 7-layer CNN and WRN-28-10 · CIFAR10 adversarial accuracy vs epsilon | adversarial accuracy | Helped |
| use global spectral norm regularization instead of standard training (ablation) | 7-layer CNN · CIFAR10 adversarial accuracy vs epsilon | adversarial accuracy | No clear effect |
| use PGA adversarial training instead of standard training | 7-layer CNN and WRN-28-10 · CIFAR10 adversarial accuracy vs epsilon | adversarial accuracy | Helped |
| add data-dependent spectral norm regularization or adversarial training | 7-layer CNN · CIFAR10 Jacobian singular values | singular value magnitude | Helped |
| add adversarial training or d.d. SNR | 7-layer CNN · CIFAR10 local linearity | deviation from linearity | Helped |
| use 10 power iterations instead of 1 for global SNR (ablation) | global spectral norm regularization · CIFAR10 | robustness | No clear effect |
Closest papers by meaning
- Regularization properties of adversarially-trained linear regression (NeurIPS 2023 · similarity 0.90)
- Adversarial Robustness through Local Linearization (NeurIPS 2019 · similarity 0.90)
- Boosting Adversarial Training via Fisher-Rao Norm-based Regularization (CVPR 2024 · similarity 0.89)
- $\sigma$-zero: Gradient-based Optimization of $\ell_0$-norm Adversarial Examples (ICLR 2025 · similarity 0.88)
- Rethinking Invariance Regularization in Adversarial Training to Improve Robustness-Accuracy Trade-off (ICLR 2025 · similarity 0.87)
- Kernel Learning with Adversarial Features: Numerical Efficiency and Adaptive Regularization (NeurIPS 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.