Research map / Adversarial robustness and security / Certified and empirical robustness
CSTAR: Towards Compact and Structured Deep Neural Networks with Adversarial Robustness
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 CSTAR low-rank decomposition instead of RAP column pruning | VGG-16 · CIFAR-10, 64× compression, Standard AT | benign accuracy: 45.08 -> 70.77 | Helped |
| Use CSTAR low-rank decomposition instead of RAP column pruning | VGG-16 · CIFAR-10, 64× compression, Standard AT | PGD-50 robust accuracy: 32.49 -> 43.13 | Helped |
| Use CSTAR low-rank decomposition instead of RAP column pruning | ResNet-18 · CIFAR-10, 16× compression, Standard AT | benign accuracy: 70.18 -> 80.90 | Helped |
| Use CSTAR low-rank decomposition instead of RAP column pruning | ResNet-18 · CIFAR-10, 16× compression, Standard AT | PGD-50 robust accuracy: 41.78 -> 47.36 | Helped |
| Use CSTAR low-rank decomposition instead of HYDRA structured pruning | VGG-16 · CIFAR-10, 10× compression, TRADES | benign accuracy: +60.99 | Helped |
| Use CSTAR low-rank decomposition instead of HYDRA structured pruning | VGG-16 · CIFAR-10, 10× compression, TRADES | robust accuracy: +28.48 | Helped |
| Use CSTAR low-rank decomposition instead of RAP | ResNet-18 · ImageNet, 16× compression, Fast AT | Top-1 accuracy: +8.58 | Helped |
| Use CSTAR low-rank decomposition instead of RAP | ResNet-18 · ImageNet, 16× compression, Fast AT | robust accuracy: +4.27 | Helped |
| Use CSTAR low-rank decomposition for structured compression | VGG-16 · CIFAR-10, 4× compression, CPU inference | inference time per image: 9.98ms -> 4.38ms/↓2.28× | Helped |
Closest papers by meaning
- Sparse DNNs with Improved Adversarial Robustness (NeurIPS 2018 · similarity 0.87)
- Architectural Adversarial Robustness: The Case for Deep Pursuit (CVPR 2021 · similarity 0.87)
- A Winning Hand: Compressing Deep Networks Can Improve Out-of-Distribution Robustness (NeurIPS 2021 · similarity 0.86)
- Holistic Adversarially Robust Pruning (ICLR 2023 · similarity 0.86)
- Adaptive Sharpness-Aware Pruning for Robust Sparse Networks (ICLR 2024 · similarity 0.86)
- Only Train Once: A One-Shot Neural Network Training And Pruning Framework (NeurIPS 2021 · similarity 0.86)
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.