Research map / Adversarial robustness and security / Certified and empirical robustness
Parseval Networks: Improving Robustness to Adversarial Examples
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 |
|---|---|---|---|
| Apply Parseval regularization (orthogonal weights and convex aggregation) | Wide ResNet-28-10 · CIFAR-10, non-adversarial training, SNR ≈40 | accuracy: 76.62 -> 81.76 | Helped |
| Apply Parseval regularization (orthogonal weights and convex aggregation) | Wide ResNet-28-10 · CIFAR-100, non-adversarial training, SNR ≈40 | accuracy: 44.62 -> 55.41 | Helped |
| Apply Parseval regularization | Wide ResNet-28-10 · CIFAR-10, clean | accuracy: 95.63 -> 96.28 | Helped |
| Apply Parseval regularization | Wide ResNet-28-10 · CIFAR-100, clean | accuracy: 79.70 -> 80.72 | Helped |
| Add convex-combination constraints to the orthogonality-constrained model (ablation) | Parseval(OC), Wide ResNet-28-10 · CIFAR-10, non-adversarial training, SNR ≈33 | accuracy: 61.38 -> 69.10 | Helped |
| Add convex-combination constraints to the orthogonality-constrained model (ablation) | Parseval(OC), Wide ResNet-28-10 · CIFAR-100, non-adversarial training, SNR ≈40 | accuracy: 49.97 -> 55.41 | Helped |
| Add convex-combination constraints to the orthogonality-constrained model (ablation) | Parseval(OC), Wide ResNet-28-10 · CIFAR-10, adversarial training, SNR ≈33 | accuracy: 85.23 -> 84.53 | Hurt |
| Combine adversarial training with Parseval regularization | Wide ResNet-28-10 · CIFAR-10, SNR ≈50 | accuracy: 93.03 -> 92.51 | Hurt |
| Combine adversarial training with Parseval regularization | Wide ResNet-28-10 · CIFAR-100, SNR ≈40 | accuracy: 55.41 -> 58.86 | Helped |
| Use orthogonality-constrained weights with convex aggregation | Wide ResNet-28-10 · SVHN, SNR ≈50 | accuracy: 97.04 -> 97.86 | Helped |
Closest papers by meaning
- Improved techniques for deterministic l2 robustness (NeurIPS 2022 · similarity 0.87)
- Towards Fast Computation of Certified Robustness for ReLU Networks (ICML 2018 · similarity 0.87)
- Evaluating Robustness of Neural Networks with Mixed Integer Programming (ICLR 2019 · similarity 0.86)
- A Recipe for Improved Certifiable Robustness (ICLR 2024 · similarity 0.86)
- Learning Better Certified Models from Empirically-Robust Teachers (NeurIPS 2026 · similarity 0.86)
- Interpreting Robustness Proofs of Deep Neural Networks (ICLR 2024 · similarity 0.85)
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.