Research map / Adversarial robustness and security / Certified and empirical robustness
A PAC-Bayes Analysis of Adversarial Robustness
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 |
|---|---|---|---|
| Train with PGDU defense | Differentiable decision-tree majority vote with signed voters · MNIST 1vs7, ℓ2, PGDU attack | averaged adversarial risk: .263 -> .021 | Helped |
| Train with PGDU defense | Differentiable decision-tree majority vote with signed voters · MNIST 1vs7, ℓ2, PGDU attack | classical adversarial risk: .245 -> .006 | Helped |
| Train with PGDU defense | Differentiable decision-tree majority vote with signed voters · MNIST 1vs7, ℓ2, IFGSMU attack | averaged adversarial risk: .066 -> .013 | Helped |
| Add uniform-noise defense | Differentiable decision-tree majority vote with signed voters · MNIST 1vs7, ℓ2, PGDU attack | averaged adversarial risk: .263 -> .151 | Helped |
| Use PGDU rather than uniform-noise defense | Differentiable decision-tree majority vote with signed voters · MNIST 1vs7, ℓ2, PGDU attack | averaged adversarial risk: .151 -> .021 | Helped |
| Use signed rather than unsigned voters (ablation) | PGDU-defended differentiable decision-tree majority vote · MNIST 1vs7, ℓ2, PGDU attack | bound: 1.389 -> .137 | Helped |
| Use signed rather than unsigned voters (ablation) | PGDU-defended differentiable decision-tree majority vote · MNIST 1vs7, ℓ2, PGDU attack | classical adversarial risk: .030 -> .028 | Helped |
Closest papers by meaning
- PAC-Bayesian Spectrally-Normalized Bounds for Adversarially Robust Generalization (NeurIPS 2023 · similarity 0.93)
- Improving Robust Generalization by Direct PAC-Bayesian Bound Minimization (CVPR 2023 · similarity 0.90)
- On the Existence of The Adversarial Bayes Classifier (NeurIPS 2021 · similarity 0.88)
- Theoretical evidence for adversarial robustness through randomization (NeurIPS 2019 · similarity 0.88)
- Fantastic Robustness Measures: The Secrets of Robust Generalization (NeurIPS 2023 · similarity 0.88)
- Towards Million-Scale Adversarial Robustness Evaluation With Stronger Individual Attacks (CVPR 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.