Research map / Optimization / Optimizers
A Variant of Anderson Mixing with Minimal Memory Size
Fuchao Wei, Chenglong Bao, Yang Liu, Guangwen Yang
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 |
|---|---|---|---|
| train with sMin-AM instead of SGDM | ResNet18 · CIFAR-10 | top-1 test accuracy: 94.82 -> 95.28 | Improved |
| train with sMin-AM instead of SGDM | VGG16 · CIFAR-10 | top-1 test accuracy: 93.52 -> 94.09 | Improved |
| train with sMin-AM instead of Adam | ResNet18 · CIFAR-100 | top-1 test accuracy: 72.41 -> 78.26 | Improved |
| use sMin-AM (one memory pair) instead of SAM(10) | ResNet56 · CIFAR-10 | top-1 test accuracy: 93.77 -> 93.82 | No clear effect |
| use sMin-AM instead of ST-AM | DenseNet121 · CIFAR-100 | top-1 test accuracy: 80.36 -> 79.88 | Hurt |
| use sMin-AM instead of ST-AM | ResNet44 · CIFAR-10 | top-1 test accuracy: 93.52 -> 93.72 | Improved |
| use sMin-AM instead of SAM(10) on memory | ResNet18 · CIFAR-10 | memory (x SGDM): 1.73 -> 1.01 | Improved |
| use sMin-AM instead of SAM(10) on per-epoch time | VGG16 · CIFAR-10 | per-epoch time (x SGDM): 2.59 -> 1.25 | Improved |
| use sMin-AM instead of ST-AM on total training time | ResNet18 · CIFAR-10 | total time (x SGDM): 0.82 -> 0.64 | Improved |
| train with sMin-AM instead of SGDM | ResNet50 · ImageNet | test accuracy: 75.81 -> 76.04 | Improved |
| train with sMin-AM instead of SGDM | ResNet50 · ImageNet | train accuracy: 77.60 -> 77.68 | Improved |
| sMin-AM overhead vs SGDM on ImageNet | ResNet50 · ImageNet | memory (x SGDM): 1.00 -> 1.04 | Hurt |
Closest papers by meaning
- A Class of Short-term Recurrence Anderson Mixing Methods and Their Applications (ICLR 2022 · similarity 0.92)
- Stochastic Anderson Mixing for Nonconvex Stochastic Optimization (NeurIPS 2021 · similarity 0.90)
- Stochastic Optimization with Arbitrary Recurrent Data Sampling (ICML 2024 · similarity 0.85)
- SGD with memory: fundamental properties and stochastic acceleration (ICLR 2025 · similarity 0.85)
- Lean and Mean Adaptive Optimization via Subset-Norm and Subspace-Momentum with Convergence Guarantees (ICML 2025 · similarity 0.84)
- Shuffling Gradient-Based Methods for Nonconvex-Concave Minimax Optimization (NeurIPS 2024 · similarity 0.84)
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.
