AI Paper Atlas

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

Prior-Guided Adversarial Initialization for Fast Adversarial Training

Xiaojun Jia, Yong Zhang, Xingxing Wei, Baoyuan Wu, Ke Ma, Jue Wang, Xiaochun Cao

ECCV 2022 · Adversarial robustness and security › Certified and empirical robustness · approach: Adversarial training refinements · 43 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
add prior-guided regularizer (ablation)FGSM-BP · CIFAR-10 ResNet18AA accuracy: 43.30 -> 47.47Improved
add prior-guided regularizer (ablation)FGSM-EP · CIFAR-10 ResNet18AA accuracy: 43.39 -> 47.94Improved
add prior-guided regularizer (ablation)FGSM-MEP · CIFAR-10 ResNet18AA accuracy: 43.99 -> 49.00Improved
add regularizer to random-init FGSM (ablation)FGSM-RS · CIFAR-10 ResNet18AA accuracy (best): 37.07 -> 46.80Improved
replace random initialization with previous-batch adversarial perturbation (ablation)FGSM-RS · CIFAR-10 ResNet18PGD-50 accuracy (last checkpoint): 00.02 -> 44.39Improved
use momentum of all previous epochs instead of previous batch as initialization (ablation)FGSM-BP w/o regularizer · CIFAR-10 ResNet18AA accuracy (best): 43.30 -> 43.99Improved
use momentum of all previous epochs instead of previous epoch as initialization (ablation)FGSM-EP with regularizer · CIFAR-10 ResNet18AA accuracy (best): 47.94 -> 49.00Improved
use FGSM-MEP instead of random-init FGSM-RSFGSM-RS · Tiny ImageNet last checkpointPGD-10 accuracy: 0.00 -> 22.02Improved
use FGSM-MEP instead of multi-step PGD-ATPGD-AT · Tiny ImageNet best checkpointPGD-10 accuracy: 20.2 -> 23.8Improved
use FGSM-BP instead of FGSM-RS at epsilon 8FGSM-RS · ImageNet ResNet50 epsilon=8PGD-10 accuracy: 0.00 -> 21.76Improved
use FGSM-MEP instead of PGD-ATPGD-AT · CIFAR-10 ResNet18AA accuracy: about 48% -> about 49%Improved
use FGSM-MEP instead of NuAT for training timeNuAT · CIFAR-10 ResNet18training time (min): 104 -> 73 minImproved

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