Research map / Adversarial robustness and security / Certified and empirical robustness
Rethinking and Improving Robustness of Convolutional Neural Networks: a Shapley Value-based Approach in Frequency Domain
Yiting Chen, Qibing Ren, Junchi Yan
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 CSA augmentation to PGD-AT | PGD-AT ResNet-18 · CIFAR10 PGD-20 | accuracy: 46.38% -> 49.42% | Improved |
| add CSA augmentation to PGD-AT | PGD-AT ResNet-18 · CIFAR10 AutoAttack | accuracy: 44.06% -> 46.56% | Improved |
| add CSA augmentation to TRADES | TRADES ResNet-18 · CIFAR10 PGD-20 | accuracy: 51.08% -> 52.06% | Improved |
| add CSA augmentation to TRADES | TRADES ResNet-18 · CIFAR100 AutoAttack | accuracy: 22.76% -> 23.73% | Improved |
| add CSA augmentation to PGD-AT | PGD-AT ResNet-18 · CIFAR10 natural | accuracy: 84.49% -> 82.91% | Hurt |
| add CSA augmentation to PGD-AT | PGD-AT ResNet-18 · CIFAR100 PGD-20 | accuracy: 21.70% -> 23.63% | Improved |
| add CSA augmentation to PGD-AT | PGD-AT WRN-28-10 · CIFAR10 PGD-20 | accuracy: 47.51% -> 50.83% | Improved |
| add CSA augmentation to TRADES | TRADES WRN-28-10 · CIFAR10 AutoAttack | accuracy: 46.65% -> 48.65% | Improved |
| add CSA augmentation to PGD-AT | PGD-AT VGG16 · CIFAR10 AutoAttack | accuracy: 42.70% -> 43.53% | Improved |
| add CSA augmentation to PGD-AT | PGD-AT VGG16 · CIFAR10 natural | accuracy: 80.06% -> 75.40% | Hurt |
| add CSA augmentation to TRADES | TRADES VGG16 · CIFAR10 PGD-20 | accuracy: 46.63% -> 48.23% | Improved |
| suppress high frequency components with TRADES (ablation) | TRADES ResNet-18 · CIFAR10 AutoAttack | accuracy: 47.74% -> 45.14% | Hurt |
Closest papers by meaning
- Towards Building More Robust Models with Frequency Bias (ICCV 2023 · similarity 0.88)
- Revisiting Visual Model Robustness: A Frequency Long-Tailed Distribution View (NeurIPS 2023 · similarity 0.88)
- Robust Models are less Over-Confident (NeurIPS 2022 · similarity 0.88)
- An Introduction to Adversarially Robust Deep Learning (TPAMI 2023 · similarity 0.88)
- A Comprehensive Study on Robustness of Image Classification Models: Benchmarking and Rethinking (IJCV 2024 · similarity 0.88)
- Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data (ICLR 2025 · similarity 0.86)
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.
