AI Paper Atlas

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

NeurIPS 2021 · Adversarial robustness and security › Certified and empirical robustness · approach: Adversarial training recipes · 0 citations (OpenAlex)

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 changedWhereMeasuredEffect
Apply GDWS post-training to pre-trained ResNet-18adversarially trained ResNet-18 · CIFAR-10FPS: 28 -> 104Improved
Apply GDWS to ResNet-18adversarially trained ResNet-18 · CIFAR-10robust accuracy: 51.55 -> 50.98No clear effect
Apply GDWS to WRN-28-4 rather than using RobNetWRN-28-4 · CIFAR-10FPS: 5 -> 65Improved
Apply GDWS to ResNet-50adversarially trained ResNet-50 · CIFAR-10FPS: 16 -> 37Improved
Apply GDWS to VGG-16adversarially trained VGG-16 · CIFAR-10FPS: 36 -> 129Improved
Apply GDWS to ResNet-18 instead of training lightweight DWS networks from scratchResNet-18 + GDWS · CIFAR-10robust accuracy: 50.98 vs 49.08Improved
Use lightweight networks instead of GDWSMobileNetV1 · CIFAR-10model size: 12.3 vs 29.1Improved
Use GDWS instead of ADMM 75% structured pruning on ResNet-18ResNet-18 baseline · CIFAR-10robust accuracy: 43.23 -> 46.15Improved
Use ADMM structured pruning (p=75%) instead of GDWSResNet-18 · CIFAR-10natural accuracy: 80.65 -> 74.62Hurt
Apply GDWS to 99% HYDRA pruned WRN-28-4HYDRA pruned WRN-28-4 · CIFAR-10FPS: 28 -> 68Improved
Apply HYDRA 99% pruning alone to WRN-28-4WRN-28-4 · CIFAR-10robust accuracy: 57.23 -> 47.21Hurt
Apply GDWS to ResNet-50 on ImageNetadversarially trained ResNet-50 · ImageNetFPS: 15 -> 19Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy