Research map / Adversarial robustness and security / Certified and empirical robustness
Consistency Regularization for Certified Robustness of Smoothed Classifiers
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 |
|---|---|---|---|
| Add consistency regularization (λ=10) to Gaussian training | ResNet-110 · CIFAR-10, σ=0.50 | ACR: 0.525 -> 0.720 | Helped |
| Add consistency regularization (λ=5) to Gaussian training | ResNet-110 · CIFAR-10, σ=1.00 | ACR: 0.542 -> 0.734 | Helped |
| Add consistency regularization (λ=1) to the searched SmoothAdv model | ResNet-110 · CIFAR-10, σ=1.00 | ACR: 0.785 -> 0.816 | Helped |
| Add consistency regularization (λ=5) to Gaussian training | ResNet-50 · ImageNet, σ=0.50 | ACR: 0.733 -> 0.822 | Helped |
| Add consistency regularization (λ=5) to Gaussian training | ResNet-50 · ImageNet, σ=1.00 | ACR: 0.875 -> 0.982 | Helped |
| Add consistency regularization to SmoothAdv | ResNet-110 · CIFAR-10, σ=0.50 | ACR: 0.717 -> 0.726 | Helped |
| Use consistency regularization (λ=5) with Gaussian training | ResNet-50 · ImageNet, σ=0.50 | ACR: 0.825 vs 0.822 | No clear effect |
| Use consistency regularization (λ=5) with Gaussian training | ResNet-50 · ImageNet, σ=1.00 | ACR: 1.040 -> 0.982 | Hurt |
| Add consistency regularization (λ=10) to Gaussian training (ablation) | ResNet-20 · CIFAR-10, σ=0.50 | ACR: 0.524 -> 0.686 | Helped |
| Add consistency regularization (λ=10) to Gaussian training (ablation) | DenseNet-40 · CIFAR-10, σ=0.50 | ACR: 0.494 -> 0.661 | Helped |
| Replace the proposed consistency loss with MSE consistency (ablation) | Gaussian-trained classifier · MNIST, σ=1.00 | ACR | Hurt |
| Remove the entropy penalty from the consistency loss (ablation) | Gaussian-trained classifier · MNIST, σ=1.00 | ACR | Hurt |
Closest papers by meaning
- Regularized Training and Tight Certification for Randomized Smoothed Classifier with Provable Robustness (AAAI 2020 · similarity 0.95)
- Confidence-Aware Training of Smoothed Classifiers for Certified Robustness (AAAI 2023 · similarity 0.94)
- DRF: Improving Certified Robustness via Distributional Robustness Framework (AAAI 2024 · similarity 0.90)
- Certified Adversarial Robustness via Randomized $\alpha$-Smoothing for Regression Models (NeurIPS 2024 · similarity 0.89)
- Mitigating the Curse of Dimensionality for Certified Robustness via Dual Randomized Smoothing (ICLR 2024 · similarity 0.89)
- Treatment of Statistical Estimation Problems in Randomized Smoothing for Adversarial Robustness (NeurIPS 2024 · similarity 0.89)
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.