Research map / Adversarial robustness and security / Certified and empirical robustness
Adversarial Robustness with Semi-Infinite Constrained Learning
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 DALE with ρ=0.8 instead of MART | CIFAR-10 ResNet-18 · CIFAR-10 PGD^20 | accuracy: 49.8 -> 50.1 | Helped |
| Use DALE with ρ=0.8 instead of MART | CIFAR-10 ResNet-18 · CIFAR-10 | clean accuracy: 78.9 -> 85.0 | Hurt |
| Use DALE with ρ=1.0 instead of MART | CIFAR-10 ResNet-18 · CIFAR-10 PGD^20 | accuracy: 49.8 -> 51.7 | Helped |
| Use DALE with ρ=1.0 instead of MART | CIFAR-10 ResNet-18 · CIFAR-10 | clean accuracy: 78.9 -> 82.1 | Helped |
| Use DALE with ρ=0.8 instead of TRADES | MNIST four-layer CNN · MNIST PGD^10 | accuracy: 94.0 -> 94.3 | Helped |
| Vary the nominal constraint level ρ from 0.8 to 1.0 | DALE · CIFAR-10 PGD^20 | accuracy: 50.1 -> 51.7 | Helped |
| Vary the nominal constraint level ρ from 0.8 to 1.0 | DALE · CIFAR-10 | clean accuracy: 85.0 -> 82.1 | Hurt |
| Add adaptive dual-variable updates instead of using a fixed nominal-loss weight (ablation) | DALE · CIFAR-10 PGD^20 | accuracy: 45.7 to 49.5 for fixed ν values; adaptive DALE 50.1 at ρ=0.8 | Helped |
| Increase the fixed nominal-loss weight from ν=0.1 to ν=1.0 (ablation) | Regularized DALE · CIFAR-10 PGD^20 | accuracy: 49.5 -> 45.8 | Hurt |
Closest papers by meaning
- Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data (ICLR 2025 · similarity 0.91)
- Convergence of Adversarial Training in Overparametrized Neural Networks (NeurIPS 2019 · similarity 0.91)
- Probabilistically Robust Learning: Balancing Average and Worst-case Performance (ICML 2022 · similarity 0.91)
- Adversarial Training and Robustness for Multiple Perturbations (NeurIPS 2019 · similarity 0.91)
- Attribute-Guided Adversarial Training for Robustness to Natural Perturbations (AAAI 2021 · similarity 0.90)
- On the Duality Between Sharpness-Aware Minimization and Adversarial Training (ICML 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.