Research map / Optimization / Implicit bias and generalization
Make Sharpness-Aware Minimization Stronger: A Sparsified Perturbation Approach
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 Fisher top-k mask with random mask (ablation) | SSAM-F · ImageNet | accuracy: 77.31 -> 77.08 | Hurt |
| drop sharpest weights instead of flattest in mask update (ablation) | SSAM-D · ImageNet | accuracy: 77.25 -> 76.68 | Hurt |
| drop random weights instead of flattest (ablation) | SSAM-D · ImageNet | accuracy: 77.25 -> 77.08 | Hurt |
| sparsify the SAM perturbation (SSAM-F 50%) instead of dense SAM | VGG11-BN · CIFAR10 | accuracy: 93.87 -> 94.03 | Helped |
| increase sparsity to 99% | SSAM-F · CIFAR10 VGG11-BN | accuracy: 94.03 -> 93.47 | Hurt |
| update the mask every step rather than fixing it (ablation) | SSAM-F · CIFAR10 ResNet18 | accuracy: 96.21 -> 96.70 | Helped |
| increase Fisher sample count N_F from 16 to 4096 (ablation) | SSAM-F · CIFAR10 ResNet18 | accuracy: 96.77 -> 96.66 | No clear effect |
| increase perturbation rho beyond 0.2 on CIFAR10 (ablation) | SAM · CIFAR10 ResNet18 | accuracy: 96.83 -> 93.16 | Hurt |
Closest papers by meaning
- A Single-Step, Sharpness-Aware Minimization is All You Need to Achieve Efficient and Accurate Sparse Training (NeurIPS 2024 · similarity 0.95)
- Sharpness-Aware Minimization: General Analysis and Improved Rates (ICLR 2025 · similarity 0.92)
- Sharpness-Aware Minimization Efficiently Selects Flatter Minima Late In Training (ICLR 2025 · similarity 0.91)
- Friendly Sharpness-Aware Minimization (CVPR 2024 · similarity 0.91)
- $\boldsymbol{\mu}\mathbf{P^2}$: Effective Sharpness Aware Minimization Requires Layerwise Perturbation Scaling (NeurIPS 2024 · similarity 0.91)
- A Universal Class of Sharpness-Aware Minimization Algorithms (ICML 2024 · similarity 0.90)
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.