AI Paper Atlas

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

Improving Robustness with Adaptive Weight Decay

Amin Ghiasi, Ali Shafahi, Reza Ardekani

NeurIPS 2023 · Adversarial robustness and security › Certified and empirical robustness · approach: Adversarial training recipes · 1 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 weight decay with adaptive weight decayPGD-7 adversarial training · CIFAR-10AutoAttack accuracy: 45.19 -> 50.03Improved
replace weight decay with adaptive weight decayPGD-7 adversarial training · CIFAR-100AutoAttack accuracy: 22.53 -> 27.15Improved
replace weight decay with adaptive weight decayPGD-7 adversarial training · Tiny ImageNetAutoAttack accuracy: 16.73 -> 19.74Improved
replace weight decay with adaptive weight decayPGD-7 adversarial training · Flowersnatural accuracy: 90.98 -> 90.39Hurt
replace weight decay with adaptive weight decayPGD-7 adversarial training · CIFAR-100weight norm: 51.32 -> 13.41Improved
replace weight decay with adaptive weight decayPGD-7 adversarial training · CIFAR-100non-adaptive total loss: 0.67 -> 1.51Hurt
add ASAM optimizer to adaptive weight decay (ablation)AWD · CIFAR-100 WRN28-10AutoAttack accuracy: 27.15 -> 29.54Improved
add ASAM optimizer to standard weight decay (ablation)Weight decay · CIFAR-100 WRN28-10AutoAttack accuracy: 22.53 -> 22.55No clear effect
use AWD + ASAM instead of TRADES etcWRN32-10 · CIFAR-10AutoAttack accuracy: 53.40 -> 54.04Improved
train AWD for 50 instead of 100 epochs (ablation)AWD + ASAM WRN32-10 · CIFAR-100AutoAttack accuracy: 29.70 -> 29.25No clear effect
train AWD for 5 epochs instead of 100 (ablation)AWD + ASAM WRN32-10 · CIFAR-100AutoAttack accuracy: 29.70 -> 11.03Hurt
use AWD with ResNet50 natural trainingResNet50 · CIFAR-100training accuracy: 98.57 -> 95.29Hurt

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