Research map / Architectures and efficiency / CNN design
CNN-Cert: An Efficient Framework for Certifying Robustness of Convolutional Neural Networks
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 CNN-Cert's convolution-aware certification instead of Fast-Lin on pure CNNs | MNIST 4-layer CNN · MNIST 4-layer CNN, ℓ∞ | certified lower bound: 0.0406 -> 0.0491 | Helped |
| Use CNN-Cert rather than Fast-Lin for certification | MNIST 4-layer CNN · MNIST 4-layer CNN, ℓ∞ | runtime (sec): 9.03 -> 2.33 | Helped |
| Use CNN-Cert-ada instead of CNN-Cert-ReLU bounds (ablation) | MNIST LeNet · MNIST LeNet, ℓ∞ | certified lower bound: 0.0113 -> 0.0120 | Helped |
| Remove pooling layers (ablation) | MNIST LeNet-style CNN · MNIST LeNet, ℓ∞ | certified lower bound: 0.0120 -> 0.0273 | Helped |
| Use CNN-Cert-ada instead of CNN-Cert-ReLU bounds (ablation) | MNIST ResNet-3 · MNIST ResNet-3, ℓ∞ | certified lower bound: 0.0179 -> 0.0202 | Helped |
| Replace ReLU with sigmoid (ablation) | MNIST LeNet · MNIST LeNet, ℓ∞ | certified lower bound: 0.0113 -> 0.0124 | Helped |
| Replace ReLU with tanh (ablation) | MNIST LeNet · MNIST LeNet, ℓ∞ | certified lower bound: 0.0113 -> 0.0170 | Helped |
| Replace ReLU with arctan (ablation) | MNIST LeNet · MNIST LeNet, ℓ∞ | certified lower bound: 0.0113 -> 0.0153 | Helped |
| Use CNN-Cert-ada rather than Dual-LP | MNIST 2-layer MLP · MNIST 2-layer MLP, ℓ∞ | certified lower bound: 0.0372 -> 0.0371 | No clear effect |
Closest papers by meaning
- Towards General Robustness Verification of MaxPool-based Convolutional Neural Networks via Tightening Linear Approximation (CVPR 2024 · similarity 0.90)
- Tightening Robustness Verification of Convolutional Neural Networks with Fine-Grained Linear Approximation (AAAI 2021 · similarity 0.90)
- Learning Better Certified Models from Empirically-Robust Teachers (NeurIPS 2026 · similarity 0.90)
- Verification of Neural Networks Against Convolutional Perturbations via Parameterised Kernels (AAAI 2025 · similarity 0.89)
- Tightening Robustness Verification of MaxPool-based Neural Networks via Minimizing the Over-Approximation Zone (CVPR 2025 · similarity 0.88)
- Robust NAS under adversarial training: benchmark, theory, and beyond (ICLR 2024 · 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.