AI Paper Atlas

Research map / Optimization / Implicit bias and generalization

The Crucial Role of Normalization in Sharpness-Aware Minimization

Yan Dai, Kwangjun Ahn, Suvrit Sra

NeurIPS 2023 · Optimization › Implicit bias and generalization · approach: Sharpness-aware minimization theory · 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
remove gradient normalization from SAMSAM · CIFAR-10 early-stage training, CNN and FCNtraining loss stabilityHurt
use normalized SAM updates instead of un-normalized updatesSAM and USAM · CIFAR-10 test, ResNet18 adversarial checkpointtest accuracyImproved
use small ρ in USAM rather than normalized SAMUSAM · CIFAR-10 test, ResNet18 adversarial checkpointtest accuracyHurt
use larger ρ in USAM rather than normalized SAMUSAM · CIFAR-10 training, ResNet18 adversarial checkpointtraining accuracy stabilityHurt
remove gradient normalization from SAMSAM · sparse coding problemtest accuracyMixed
vary ρ in USAM, using a tiny ρ=0.01 (ablation)USAM · sparse coding problemtest accuracyNo clear effect
vary ρ in USAM, using ρ=0.1 or ρ=0.3 (ablation)USAM · sparse coding problemtest accuracyImproved
vary ρ in USAM, using ρ=0.5 (ablation)USAM · sparse coding problemtraining stabilityHurt
vary η while fixing ρ (ablation)SAM · sparse coding problemtest accuracyImproved
vary η while fixing ρ (ablation)USAM · sparse coding problemtest accuracyMixed

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