Research map / Adversarial robustness and security / Certified and empirical robustness
Generalized Depthwise-Separable Convolutions for Adversarially Robust and Efficient Neural Networks
Hassan Dbouk, Naresh Shanbhag
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 |
|---|---|---|---|
| Apply GDWS post-training to pre-trained ResNet-18 | adversarially trained ResNet-18 · CIFAR-10 | FPS: 28 -> 104 | Improved |
| Apply GDWS to ResNet-18 | adversarially trained ResNet-18 · CIFAR-10 | robust accuracy: 51.55 -> 50.98 | No clear effect |
| Apply GDWS to WRN-28-4 rather than using RobNet | WRN-28-4 · CIFAR-10 | FPS: 5 -> 65 | Improved |
| Apply GDWS to ResNet-50 | adversarially trained ResNet-50 · CIFAR-10 | FPS: 16 -> 37 | Improved |
| Apply GDWS to VGG-16 | adversarially trained VGG-16 · CIFAR-10 | FPS: 36 -> 129 | Improved |
| Apply GDWS to ResNet-18 instead of training lightweight DWS networks from scratch | ResNet-18 + GDWS · CIFAR-10 | robust accuracy: 50.98 vs 49.08 | Improved |
| Use lightweight networks instead of GDWS | MobileNetV1 · CIFAR-10 | model size: 12.3 vs 29.1 | Improved |
| Use GDWS instead of ADMM 75% structured pruning on ResNet-18 | ResNet-18 baseline · CIFAR-10 | robust accuracy: 43.23 -> 46.15 | Improved |
| Use ADMM structured pruning (p=75%) instead of GDWS | ResNet-18 · CIFAR-10 | natural accuracy: 80.65 -> 74.62 | Hurt |
| Apply GDWS to 99% HYDRA pruned WRN-28-4 | HYDRA pruned WRN-28-4 · CIFAR-10 | FPS: 28 -> 68 | Improved |
| Apply HYDRA 99% pruning alone to WRN-28-4 | WRN-28-4 · CIFAR-10 | robust accuracy: 57.23 -> 47.21 | Hurt |
| Apply GDWS to ResNet-50 on ImageNet | adversarially trained ResNet-50 · ImageNet | FPS: 15 -> 19 | Improved |
Closest papers by meaning
- Robustness and Generalization via Generative Adversarial Training (ICCV 2021 · similarity 0.87)
- A Combination of Noise and Bilateral Filters Achieve Supralinear and Scalable Adversarial Robustness in CNNs (CVPR 2026 · similarity 0.87)
- Learning Extremely Lightweight and Robust Model with Differentiable Constraints on Sparsity and Condition Number (ECCV 2022 · similarity 0.86)
- Confidence Propagation through CNNs for Guided Sparse Depth Regression (TPAMI 2019 · similarity 0.86)
- Sparsity Winning Twice: Better Robust Generalization from More Efficient Training (ICLR 2022 · similarity 0.86)
- The Master Key Filters Hypothesis: Deep Filters Are General (AAAI 2025 · 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.
