Research map / Adversarial robustness and security / Certified and empirical robustness
Towards Understanding the Regularization of Adversarial Robustness on Neural Networks
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 |
|---|---|---|---|
| Increase adversarial training strength from 4 to 16 (ablation) | ResNet-56 · CIFAR10 test | accuracy: 89.32 -> 82.83 | Hurt |
| Increase adversarial training strength from 4 to 16 (ablation) | ResNet-56 · CIFAR100 test | accuracy: 62.01 -> 56.30 | Hurt |
| Increase adversarial training strength from 4 to 16 (ablation) | ResNet-56 · CIFAR10 | loss gap: 0.429 -> 0.385 | Helped |
| Increase adversarial training strength from 4 to 16 (ablation) | ResNet-56 · CIFAR100 | loss gap: 1.307 -> 0.975 | Helped |
| Increase adversarial training strength from 4 to 16 (ablation) | ResNet-56 · CIFAR10 | training loss: 0.038 -> 0.252 | Hurt |
| Increase adversarial training strength from 4 to 16 (ablation) | ResNet-56 · CIFAR100 | training loss: 0.469 -> 0.822 | Hurt |
| Increase adversarial training strength from 4 to 16 (ablation) | ResNet-56 · CIFAR10 test | test loss: 0.467 -> 0.637 | Hurt |
| Increase adversarial training strength from 4 to 16 (ablation) | ResNet-56 · CIFAR100 test | test loss: 1.776 -> 1.797 | Mixed |
| Use stronger adversarial training (ablation) | ResNet-56 · CIFAR10 | average maximal loss variation: 0.1337±0.2494 -> 0.0888±0.1816 | Helped |
| Use stronger adversarial training (ablation) | ResNet-56 · CIFAR100 | average maximal loss variation: 0.1868±0.2946 -> 0.1355±0.2494 | Helped |
| Replace PGD adversarial training with FGSM adversarial training (ablation) | ResNet-56 · CIFAR10/100 | standard test accuracy | No clear effect |
Closest papers by meaning
- Improving the Adversarial Robustness and Interpretability of Deep Neural Networks by Regularizing Their Input Gradients (AAAI 2018 · similarity 0.92)
- Rethinking Invariance Regularization in Adversarial Training to Improve Robustness-Accuracy Trade-off (ICLR 2025 · similarity 0.90)
- Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data (ICLR 2025 · similarity 0.90)
- Kernel Learning with Adversarial Features: Numerical Efficiency and Adaptive Regularization (NeurIPS 2025 · similarity 0.89)
- The Price of Implicit Bias in Adversarially Robust Generalization (NeurIPS 2024 · similarity 0.89)
- Stability and Generalization of Adversarial Training for Shallow Neural Networks with Smooth Activation (NeurIPS 2024 · similarity 0.89)
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.