Research map / Adversarial robustness and security / Certified and empirical robustness
Non-Parametric Probabilistic Robustness: A Conservative Risk Estimator under Unknown Perturbation Distributions
Zheng Wang, Yi Zhang, Siddartha Khastgir, carsten maple, Xingyu Zhao
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 learnable up-sampler to base GMM (K=3), label dependence (ablation) | NPPR estimator · CIFAR-10 ResNet18 | NPPR (%), lower is more conservative: 75.79 -> 69.08 | Improved |
| add learnable up-sampler, input-independent setting (ablation) | NPPR estimator · CIFAR-10 ResNet18 | NPPR (%): 74.99 -> 81.27 | Hurt |
| add learnable up-sampler, joint dependence (ablation) | NPPR estimator · CIFAR-10 ResNet18 | NPPR (%): 79.20 -> 78.59 | Improved |
| increase mixture components from K=3 to K=12, joint dependence (ablation) | NPPR estimator · CIFAR-10 ResNet18 | NPPR (%): 78.59 -> 73.11 | Improved |
| increase mixture components from K=7 to K=12, input-independent (ablation) | NPPR estimator · CIFAR-10 ResNet18 | NPPR (%): 73.25 -> 74.43 | Hurt |
| make perturbation distribution label dependent rather than independent (K=7, 16/255) (ablation) | NPPR estimator · CIFAR-10 ResNet18 | NPPR (%): 73.25 -> 64.88 | Improved |
| make perturbation input dependent rather than independent (K=7, 16/255) (ablation) | NPPR estimator · CIFAR-10 ResNet18 | NPPR (%): 73.25 -> 84.17 | Hurt |
| replace fixed Uniform perturbations with learned NPPR distribution | PR evaluation · CIFAR-10 ResNet18 (16/255) | robustness (%): 97.26 -> 76.27 | Improved |
| train estimator at larger radius (16/255 vs 4/255), evaluate at 32/255 (ablation) | NPPR estimator · CIFAR-10 ResNet18 | NPPR (%): 59.79 -> 52.49 | Improved |
| use ResNet50 as GMM feature extractor instead of ResNet18, evaluating ResNet18 (ablation) | NPPR estimator · CIFAR-10 (L_inf 16/255) | NPPR (%): 76.27 -> 86.41 | Hurt |
Closest papers by meaning
- Probabilistically Robust Learning: Balancing Average and Worst-case Performance (ICML 2022 · similarity 0.90)
- Adversarial Training for Probabilistic Robustness (ICCV 2025 · similarity 0.90)
- Adversarial Robustness with Semi-Infinite Constrained Learning (NeurIPS 2021 · similarity 0.89)
- Learning perturbation sets for robust machine learning (ICLR 2021 · similarity 0.88)
- Consistent Adversarially Robust Linear Classification: Non-Parametric Setting (ICML 2024 · similarity 0.87)
- Probably Approximately Global Robustness Certification (ICML 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.
