Research map / Adversarial robustness and security / Certified and empirical robustness
Fastened CROWN: Tightened Neural Network Robustness Certificates
Zhaoyang Lyu, Ching-Yun Ko, Zhifeng Kong, Ngai Wong, Dahua Lin, Luca Daniel
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 |
|---|---|---|---|
| Optimize the activation bounding-line parameters used by CROWN | CROWN · Sensorless Drive Diagnosis, l∞, 12 × [20] Sigmoid | certified l∞ bound: 0.07820402 -> 0.15097938 | Improved |
| Optimize the activation bounding-line parameters used by CROWN | CROWN · MNIST, l1, 5 × [100] Sigmoid | certified l1 bound: 2.09984541 -> 2.51844215 | Improved |
| Optimize the activation bounding-line parameters used by CROWN | CROWN · CIFAR10, l∞, 8 × [2048] Sigmoid | certified l∞ bound: 0.00237977 -> 0.00327576 | Improved |
| Optimize the activation bounding-line parameters used by CROWN | CROWN · CIFAR10, l2, 6 × [2048] ReLU | certified l2 bound: 0.37454715 -> 0.37229005 | Hurt |
| Use LP-based bounds with multiple bounding lines instead of FROWN | FROWN · Sensorless Drive Diagnosis, l∞, 4 × [20] ReLU | certified l∞ bound: 0.22465439 -> 0.22689147 | Improved |
| Use LP-based bounds with multiple bounding lines instead of FROWN | FROWN · Sensorless Drive Diagnosis, l∞, 12 × [20] Sigmoid | certified l∞ bound: 0.15097938 -> 0.10809507 | Hurt |
Closest papers by meaning
- A Convex Relaxation Barrier to Tight Robustness Verification of Neural Networks (NeurIPS 2019 · similarity 0.88)
- Beta-CROWN: Efficient Bound Propagation with Per-neuron Split Constraints for Neural Network Robustness Verification (NeurIPS 2021 · similarity 0.88)
- Second-Order Provable Defenses against Adversarial Attacks (ICML 2020 · similarity 0.87)
- Towards Evaluating and Training Verifiably Robust Neural Networks (CVPR 2021 · similarity 0.87)
- Efficient Neural Network Robustness Certification with General Activation Functions (NeurIPS 2018 · similarity 0.87)
- Towards Trustworthy Deep Learning (AAAI 2024 · similarity 0.85)
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.
