Research map / Optimization / Implicit bias and generalization
Understanding SAM through Minimax Perspective
Ying Chen, Aoxi Li, Javad Lavaei
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add multiple inner ascent steps (M=5 to M=10) in the final phase at rho 0.2 (ablation) | Periodic MSAM · CIFAR-100, ResNet-20 | test accuracy: 69.76% -> 70.32% | Improved |
| use larger final-phase radius (0.2 instead of 0.1) with M=10 (ablation) | Periodic MSAM · CIFAR-100, ResNet-20 | test accuracy: 70.01% -> 70.32% | Improved |
| replace standard SAM with Periodic MSAM (M=10, rho 0.2 in last 20% epochs) | SAM · CIFAR-100, ResNet-20 | test accuracy: 69.32% -> 70.32% | Improved |
| replace SAM with Periodic MSAM (rho 0.2, M=5) in last 20% of training | SAM rho 0.1 · Imagenette | top-1 validation accuracy: 75.01% -> 76.87% | Improved |
| use multi-step MSAM (M=5) instead of single-step SAM at rho 0.10 | SAM · CIFAR-100, ResNet56 | test accuracy: 74.2 -> 74.8 | Improved |
| use multi-step MSAM (M=5) instead of SAM at rho 0.05 | SAM · CIFAR-100, ResNet56 | test accuracy: 75.2 -> 75.2 | No clear effect |
| replace SGD with SAM/MSAM | SGD · CIFAR-100, ResNet56 rho 0.10 | test accuracy: 74.4 -> 74.2 (SAM), 74.8 (MSAM) | Mixed |
| replace SGD with MSAM (M=3, rho 0.10) | SGD · CIFAR-10, ResNet20 | test accuracy: 93.2 -> 93.7 | Improved |
| increase inner steps from M=5 to M=10 at small rho 0.025 (ablation) | MSAM · CIFAR-10, ResNet20 | test accuracy: 93.6 -> 93.3 | Hurt |
| use MSAM (M=5) instead of standard SAM at rho 0.10 | SAM · CIFAR-10, WRN28-2 | test accuracy: 95.6 -> 95.8 | Improved |
| use MSAM (M=5) instead of SAM at rho 0.05 | SAM · CIFAR-10, ResNet56 | test accuracy: 95.3 -> 95.2 | No clear effect |
| replace single-step SAM with MSAM (M=5) at large rho 0.4 | SAM · Matrix completion n=100 | final loss: 1.5027 -> 7.5121e-05 | Improved |
Closest papers by meaning
- Improving SAM Requires Rethinking its Optimization Formulation (ICML 2024 · similarity 0.91)
- Sharpness-Aware Minimization Efficiently Selects Flatter Minima Late In Training (ICLR 2025 · similarity 0.90)
- A Universal Class of Sharpness-Aware Minimization Algorithms (ICML 2024 · similarity 0.90)
- Explicit Eigenvalue Regularization Improves Sharpness-Aware Minimization (NeurIPS 2024 · similarity 0.90)
- Lookbehind-SAM: k steps back, 1 step forward (ICML 2024 · similarity 0.90)
- Unveiling m-Sharpness Through the Structure of Stochastic Gradient Noise (NeurIPS 2025 · 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.
