AI Paper Atlas

Research map / Adversarial robustness and security / Certified and empirical robustness

Towards Defending Multiple $$\ell _p$$-Norm Bounded Adversarial Perturbations via Gated Batch Normalization

Aishan Liu, Shiyu Tang, Xinyun Chen, Lei Huang, Haotong Qin, Xianglong Liu, Dacheng Tao

IJCV 2023 · Adversarial robustness and security › Certified and empirical robustness · approach: Theory of adversarial risk · 36 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
use Conv gate in first GBN and FC gates in other GBN blocks instead of Conv gate everywhere (ablation)GBN VGG-16 · CIFAR-10PGD-l1 robust accuracy (%): 30.4 -> 59.6Improved
use Conv gate in first GBN and FC gates elsewhere instead of FC gate everywhere (ablation)GBN VGG-16 · CIFAR-10PGD-l1 robust accuracy (%): 35.1 -> 59.6Improved
use Conv gate in first GBN and FC gates elsewhere instead of Conv gate everywhere (ablation)GBN VGG-16 · CIFAR-10clean accuracy (%): 41.3 -> 83.6Improved
use Conv gate in first GBN and FC gates elsewhere instead of Conv gate everywhere (ablation)GBN VGG-16 · CIFAR-10PGD-linf robust accuracy (%): 20.2 -> 58.1Improved
use soft-gated weighting instead of hard top-1 gate (ablation)GBN ResNet-20 · CIFAR-10PGD-l1 robust accuracy (%): 57.8 -> 58.1Improved
use soft-gated weighting instead of hard top-1 gate (ablation)GBN ResNet-20 · CIFAR-10clean accuracy (%): 79.8 -> 80.2Improved
replace vanilla BN with GBNResNet-20 · CIFAR-10PGD-l1 robust accuracy (%): 0.1 -> 58.1Improved
replace vanilla BN with GBNResNet-20 · CIFAR-10clean accuracy (%): 89.4 -> 80.2Hurt
replace vanilla BN with GBNResNet-34 · Tiny-ImageNetclean accuracy (%): 54.0 -> 43.2Hurt
replace MBN (2-way gated multiple BN) with GBNResNet-34 · Tiny-ImageNetall-attacks robust accuracy (%): 18.3 -> 37.7Improved
replace MBN with GBNResNet-20 · CIFAR-10all-attacks robust accuracy (%): 20.0 -> 48.2Improved
replace MSD with GBNLeNet · MNISTall-attacks robust accuracy (%): 38.1 -> 68.5Improved

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