Research map / Optimization / Implicit bias and generalization
Gradient Norm Aware Minimization Seeks First-Order Flatness and Improves Generalization
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 |
|---|---|---|---|
| Add GAM to SGD | ResNet18 · CIFAR-10 (Basic augmentation) | accuracy: 95.32±0.13 -> 96.17±0.21 | Helped |
| Add GAM to SGD | ResNet18 · CIFAR-100 (Basic augmentation) | accuracy: 78.32±0.32 -> 79.53±0.30 | Helped |
| Add GAM to SAM | ResNet18 · CIFAR-100 (Basic augmentation) | accuracy: 79.27±0.16 -> 80.45±0.25 | Helped |
| Add GAM to the base optimizer | ResNet50 · ImageNet Top-1 | accuracy: 76.01±0.19 -> 76.59±0.15 | Helped |
| Add GAM to SAM | ResNet50 · ImageNet Top-1 | accuracy: 76.47±0.11 -> 76.86±0.15 | Helped |
| Add GAM during transfer fine-tuning | EfficientNet-b0 with SGD · Stanford Cars | accuracy: 82.14 -> 83.50 | Helped |
| Add GAM during transfer fine-tuning | Swin-t with AdamW · Stanford Cars | accuracy: 83.50 -> 84.90 | Helped |
| Add GAM to training | ResNet-18 with SGD · CIFAR-100 Hessian spectrum at epoch 200 | maximum Hessian eigenvalue: 20.83 -> 2.39 | Helped |
| Add GAM to training | ResNet-18 with SAM · CIFAR-100 Hessian spectrum at epoch 200 | Hessian trace: 541.98 -> 366.03 | Helped |
| Apply GAM in only a fraction of training iterations | SGD and SAM · CIFAR-10 and CIFAR-100 | accuracy | Helped |
Closest papers by meaning
- Sharpness-Aware Minimization Efficiently Selects Flatter Minima Late In Training (ICLR 2025 · similarity 0.91)
- Tilted Sharpness-Aware Minimization (ICML 2025 · similarity 0.91)
- Explicit Eigenvalue Regularization Improves Sharpness-Aware Minimization (NeurIPS 2024 · similarity 0.90)
- Improving Sharpness-Aware Minimization by Lookahead (ICML 2024 · similarity 0.89)
- Sharpness-Aware Minimization Enhances Feature Quality via Balanced Learning (ICLR 2024 · similarity 0.89)
- A Universal Class of Sharpness-Aware Minimization Algorithms (ICML 2024 · similarity 0.89)
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.