Research map / Adversarial robustness and security / Certified and empirical robustness
Measuring Neural Net Robustness with Constraints
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 ALP rather than L-BFGS-B to estimate pointwise robustness | LeNet · MNIST test set | adversarial frequency estimate (%): 1.32 -> 7.15 | Helped |
| Fine-tune with baseline-generated adversarial examples (ablation) | LeNet · MNIST test set | adversarial frequency under ALP estimate (%): 7.15 -> 6.89 | No clear effect |
| Fine-tune with baseline-generated adversarial examples for two rounds (ablation) | LeNet · MNIST test set | adversarial frequency under ALP estimate (%): 7.15 -> 6.97 | No clear effect |
| Fine-tune with ALP-generated adversarial examples (ablation) | LeNet · MNIST test set | adversarial frequency under ALP estimate (%): 7.15 -> 5.40 | Helped |
| Fine-tune with ALP-generated adversarial examples for two rounds (ablation) | LeNet · MNIST test set | adversarial frequency under ALP estimate (%): 7.15 -> 5.03 | Helped |
| Fine-tune with ALP-generated adversarial examples (ablation) | NiN · CIFAR-10 test set | adversarial frequency (%): 61.5% -> 59.6% | Helped |
| Fine-tune with ALP-generated adversarial examples (ablation) | NiN · CIFAR-10 test set | accuracy (%): 91.31% -> 90.35% | Hurt |
Closest papers by meaning
- Evaluating the Robustness of Neural Networks: An Extreme Value Theory Approach (ICLR 2018 · similarity 0.91)
- Evaluating Robustness of Neural Networks with Mixed Integer Programming (ICLR 2019 · similarity 0.90)
- A Statistical Approach to Assessing Neural Network Robustness (ICLR 2019 · similarity 0.90)
- Interpreting Robustness Proofs of Deep Neural Networks (ICLR 2024 · similarity 0.89)
- Probably Approximately Global Robustness Certification (ICML 2025 · similarity 0.88)
- Learning Better Certified Models from Empirically-Robust Teachers (NeurIPS 2026 · similarity 0.88)
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.