AI Paper Atlas

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

Revisiting Outer Optimization in Adversarial Training

Ali Dabouei, Fariborz Taherkhani, Sobhan Soleymani, Nasser M. Nasrabadi

ECCV 2022 · Adversarial robustness and security › Certified and empirical robustness · approach: Provable robust learning guarantees · 4 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
replace MSGD with ENGM as outer optimizervanilla AT · CIFAR-10, ResNet-18AutoAttack accuracy: 46.77 -> 49.24Improved
replace MSGD with ENGM as outer optimizervanilla AT · CIFAR-10, ResNet-18robust overfitting: 13.2 -> 3.9Improved
replace MSGD with ENGM as outer optimizervanilla AT · CIFAR-10, ResNet-18natural accuracy: 84.70 -> 83.44Hurt
add gradient norm clipping to MSGD (ablation)MSGD · CIFAR-10, ResNet-18AutoAttack accuracy: 46.77 -> 47.59Improved
replace per-example bounding with normalizing to constant (F-ENGM) (ablation)ENGM · CIFAR-10, ResNet-18AutoAttack accuracy: 49.24 -> 46.54Hurt
use ENGM instead of naive approximation N-ENGM (ablation)ENGM · CIFAR-10, ResNet-18AutoAttack accuracy: 48.06 -> 49.24Improved
use ENGM with exact norms instead of A-ENGM (ablation)ENGM · CIFAR-10, ResNet-18execution time (sec./iter): 0.83 -> 5.06Hurt
replace MSGD with ENGM on TRADESTRADES · CIFAR-100, ResNet-18AutoAttack accuracy: 26.93 -> 29.20Improved
replace MSGD with ENGM on AWPAWP · Tiny-ImageNet, ResNet-18AutoAttack accuracy: 13.09 -> 16.02Improved
replace MSGD with ENGM on vanilla ATvanilla AT · Tiny-ImageNet, ResNet-18natural accuracy: 35.71 -> 29.78Hurt
replace MSGD with ENGM on AWPAWP · CIFAR-10, WRN-34-10AutoAttack accuracy: 56.17 -> 57.45Improved
replace MSGD with N-ENGM on AWPAWP · CIFAR-10, WRN-34-10AutoAttack accuracy: 56.17 -> 57.11Improved

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