Research map / Optimization / Optimizers
Gradient Centralization: A New Optimization Technique for Deep Neural Networks
Hongwei Yong, Jianqiang Huang, Xiansheng Hua, Lei Zhang
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 gradient centralization | SGDM, ResNet18 · CIFAR100 | accuracy: 76.87 -> 78.82 | Improved |
| add gradient centralization | SGDM, VGG11 · CIFAR100 | accuracy: 70.94 -> 71.69 | Improved |
| add gradient centralization | Adam, ResNet50 · CIFAR100 | accuracy: 71.64 -> 72.80 | Improved |
| add gradient centralization | SGDW, ResNet50 · CIFAR100 | accuracy: 74.02 -> 76.82 | Improved |
| add gradient centralization | SGDM weight decay 0, ResNet50 · CIFAR100 | accuracy: 71.62 -> 72.83 | Improved |
| add gradient centralization | SGDM weight decay 1e-4, ResNet50 · CIFAR100 | accuracy: 73.91 -> 76.56 | Improved |
| add gradient centralization | SGDM lr 0.2, ResNet50 · CIFAR100 | accuracy: 76.53 -> 77.71 | Improved |
| add gradient centralization | Adam lr 0.0015, ResNet50 · CIFAR100 | accuracy: 70.63 -> 71.22 | Improved |
| add gradient centralization | ResNet50 with GN · ImageNet val | top-1 error: 24.50 -> 23.53 | Improved |
| add gradient centralization | ResNet101 with GN · ImageNet val | top-1 error: 23.34 -> 22.14 | Improved |
| add gradient centralization | ResNet50 fine-tuning · Stanford Cars | accuracy: 88.66 -> 90.03 | Improved |
| add gradient centralization | ResNet50 fine-tuning · Stanford Dogs | accuracy: 76.16 -> 78.23 | Improved |
Closest papers by meaning
- Penalizing Gradient Norm for Efficiently Improving Generalization in Deep Learning (ICML 2022 · similarity 0.89)
- Scalable and Practical Natural Gradient for Large-Scale Deep Learning (TPAMI 2020 · similarity 0.88)
- Learning Gradient Descent: Better Generalization and Longer Horizons (ICML 2017 · similarity 0.88)
- Do Neural Networks Need Gradient Descent to Generalize? A Theoretical Study (NeurIPS 2025 · similarity 0.87)
- Optimal Rates for Generalization of Gradient Descent for Deep ReLU Classification (NeurIPS 2025 · similarity 0.86)
- Changing the Training Data Distribution to Reduce Simplicity Bias Improves In-distribution Generalization (NeurIPS 2024 · similarity 0.86)
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.
