Research map / Learning theory and algorithms / Combinatorial and search optimization
Robustness Verification of Classification Deep Neural Networks via Linear Programming
Check 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 RobustVerifier instead of RobustLU | MNIST robustness verification · MNIST, 5×5 top-left perturbation block, ε=0.10–0.30 | recognition rate: 95.94% vs 92.56%; 92.16% vs 84.57%; 87.90% vs 83.88%; 79.43% vs 67.74%; 70.56% vs 65.32% | Helped |
| Use RobustVerifier instead of GNEVER | MNIST robustness verification · MNIST, 5×5 top-left perturbation block, ε=0.10–0.30 | recognition rate: 95.94% vs 90.39%; 92.16% vs 82.00%; 87.90% vs 64.65%; 79.43% vs 65.78%; 70.56% vs 45.74% | Helped |
| Increase perturbation radius | RobustVerifier · MNIST, 5×5 top-left perturbation block | recognition rate: 95.94% -> 70.56% | Hurt |
| Use RobustVerifier instead of RobustLU | MNIST robustness verification · MNIST, ε=0.20, 5–10 layers | recognition rate: 87.90% vs 83.88%; 72.33% vs 51.84%; 61.33% vs 43.11%; 50.58% vs 29.02%; 45.90% vs 33.28%; 34.65% vs 14.19% | Helped |
| Use RobustVerifier instead of GNEVER | MNIST robustness verification · MNIST, ε=0.20, 5–10 layers | recognition rate: 87.90% vs 64.65%; 72.33% vs 50.24%; 61.33% vs 29.07%; 50.58% vs 27.79%; 45.90% vs 31.63%; 34.65% vs 12.40% | Helped |
| Increase the number of network layers | RobustVerifier · MNIST, ε=0.20 | recognition rate: 87.90% -> 34.65% | Hurt |
| Use RobustVerifier rather than RobustLU or GNEVER | GTSRB CDNN · GTSRB, 3×3 central block, ε=0.05 on one RGB channel | successful robustness verification | Helped |
| Enlarge the perturbation radius or block size | RobustVerifier · GTSRB CDNN | successful robustness verification | Hurt |
| Increase perturbation radius from ε=0.05 to ε=0.1 | GTSRB CDNN · GTSRB, 3×3 central block, one RGB channel | robustness | Hurt |
| Use RobustVerifier rather than GNEVER | Caltech-101 CDNN · Caltech-101, 40×40 central block, ε=0.5 | successful verification rate: more than 82%; 23% and 81% | Helped |
Closest papers by meaning
- Tightening Robustness Verification of Convolutional Neural Networks with Fine-Grained Linear Approximation (AAAI 2021 · similarity 0.89)
- Tightening Robustness Verification of MaxPool-based Neural Networks via Minimizing the Over-Approximation Zone (CVPR 2025 · similarity 0.88)
- Towards General Robustness Verification of MaxPool-based Convolutional Neural Networks via Tightening Linear Approximation (CVPR 2024 · similarity 0.87)
- Interpreting Robustness Proofs of Deep Neural Networks (ICLR 2024 · similarity 0.87)
- Probably Approximately Global Robustness Certification (ICML 2025 · similarity 0.86)
- VNN: Verification-Friendly Neural Networks with Hard Robustness Guarantees (ICML 2024 · similarity 0.86)
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.