Research map / Optimization / Implicit bias and generalization
Efficient Sharpness-aware Minimization for Improved Training of Neural Networks
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 changed | Where | Measured | Effect |
|---|---|---|---|
| Replace SAM with ESAM, combining stochastic weight perturbation and sharpness-sensitive data selection | SAM · CIFAR-100, PyramidNet-110 | accuracy: 84.46 -> 85.56 | Helped |
| Replace SAM with ESAM, combining stochastic weight perturbation and sharpness-sensitive data selection | SAM · CIFAR-10, ResNet-18 | images/s: 1,717 -> 2,409 | Helped |
| Replace SAM with ESAM | SAM · ImageNet, ResNet-50 | accuracy: 76.70* -> 77.05 | Helped |
| Replace SAM with ESAM | SAM · ImageNet, ResNet-101 | accuracy: 78.60* -> 79.09 | Helped |
| Add SWP to SAM (ablation) | SAM · CIFAR-10, ResNet-18 | accuracy: 96.52 -> 96.74 | Helped |
| Add SWP to SAM (ablation) | SAM · CIFAR-100, ResNet-18 | accuracy: 80.17 -> 80.53 | Helped |
| Add SDS to SAM (ablation) | SAM · CIFAR-10, ResNet-18 | accuracy: 96.52 -> 96.45 | Hurt |
| Add SDS to SAM (ablation) | SAM · CIFAR-100, ResNet-18 | accuracy: 80.17 -> 80.38 | Helped |
| Use SWP only (ablation) | SAM · ImageNet, ViT-S/16 | accuracy: 80.34 -> 80.88 | Helped |
| Use SDS only (ablation) | SAM · ImageNet, ViT-S/16 | accuracy: 80.34 -> 79.97 | Hurt |
Closest papers by meaning
- Friendly Sharpness-Aware Minimization (CVPR 2024 · similarity 0.93)
- Sharpness-Aware Minimization Efficiently Selects Flatter Minima Late In Training (ICLR 2025 · similarity 0.93)
- Sharpness-Aware Minimization: General Analysis and Improved Rates (ICLR 2025 · similarity 0.93)
- A Universal Class of Sharpness-Aware Minimization Algorithms (ICML 2024 · similarity 0.92)
- Sharpness-Aware Minimization Enhances Feature Quality via Balanced Learning (ICLR 2024 · similarity 0.91)
- A Single-Step, Sharpness-Aware Minimization is All You Need to Achieve Efficient and Accurate Sparse Training (NeurIPS 2024 · similarity 0.91)
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.