AI Paper Atlas

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

Probabilistic Margins for Instance Reweighting in Adversarial Training

Qizhou Wang, Feng Liu, Bo Han, Tongliang Liu, Chen Gong, Gang Niu, Mingyuan Zhou, Masashi Sugiyama

NeurIPS 2021 · Adversarial robustness and security › Certified and empirical robustness · approach: Adversarial training refinements · 21 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 LPS-based reweighting with PM-based reweightingAT, MART, and TRADES · CIFAR-10AA accuracy: 0.70-7.93%Improved
Replace LPS-based reweighting with PM-based reweightingAT, MART, and TRADES · CIFAR-10PGD accuracy: 0.14-2.00%Improved
Use adversarial PM rather than natural PMMAIL · CIFAR-10AA accuracy: 44.25 \pm 0.45 vs 44.10 \pm 0.21; 50.60 \pm 0.22 vs 49.52 \pm 0.11Mixed
Use normalized probability-space margin instead of logit-space multi-class margin (ablation)TRADES with MAIL · CIFAR-10AA accuracy: 49.20 \pm 0.09 -> 50.60 \pm 0.22Improved
Use sigmoid rather than hinge weight assignment (ablation)TRADES with MAIL · CIFAR-10AA accuracy: 43.40 \pm 0.22 -> 50.60 \pm 0.22Improved
Use sigmoid rather than step weight assignment (ablation)TRADES with MAIL · CIFAR-10AA accuracy: 50.30 \pm 0.20 -> 50.60 \pm 0.22Improved
Use MAIL-AT rather than standard ATResNet-18 · CIFAR-10PGD accuracy: 48.91 \pm 0.14 -> 55.25 \pm 0.23Improved
Use MAIL-TRADES rather than standard TRADESResNet-18 · CIFAR-10AA accuracy: 48.10 \pm 0.26 -> 50.60 \pm 0.22Improved
Use MAIL-AT rather than GAIRATResNet-18 · CIFAR-10CW accuracy: 39.86 \pm 0.08 -> 48.88 \pm 0.11Improved
Choose MAIL-AT rather than MAIL-TRADESMAIL · CIFAR-10, ResNet-18PGD accuracy: 53.68 \pm 0.14 -> 55.25 \pm 0.23Improved
Choose MAIL-TRADES rather than MAIL-ATMAIL · CIFAR-10, ResNet-18AA accuracy: 44.22 \pm 0.21 -> 50.60 \pm 0.22Improved
Use MAIL-AT rather than GAIRATResNet-18 · SVHNAA accuracy: 37.00 \pm 0.30 -> 41.12 \pm 0.30Improved

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