Research map / Architectures and efficiency / RNNs and normalization
Centered Weight Normalization in Accelerating Training of Deep Neural Networks
Lei Huang, Xianglong Liu, Yang Liu, Bo Lang, Dacheng Tao
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 weight centering (CWN) to weight normalization (WN) (ablation) | WN · Yale-B (MLP) | test error: 4.68 -> 4.20 | Improved |
| add weight centering (CWN) to weight normalization (WN) (ablation) | WN · SVHN permutation-invariant (MLP) | test error: 17.12 -> 16.16 | Improved |
| CWN vs plain with He-Init | plain · SVHN permutation-invariant (MLP, He-Init) | test error: 18.35 -> 16.71 | Improved |
| CWN vs plain with Xavier-Init | plain · Yale-B (MLP, Xavier-Init) | test error: 6.47 -> 3.84 | Improved |
| add WN to plain network | plain · CIFAR-10 (ResNet-56) | test error: 7.34 -> 7.58 | Hurt |
| add CWN to plain ResNet-56 | plain · CIFAR-100 (ResNet-56) | test error: 29.38 -> 29.23 | No clear effect |
| add WN to plain VGG-A | plain · CIFAR-10 (VGG-A) | test error: 14.33% -> 15.89% | Hurt |
| add CWN to plain VGG-A | plain · CIFAR-10 (VGG-A) | test error: 14.33% -> 13.06% | Improved |
| add CWN to plain GoogLeNet | plain · CIFAR-100 (GoogLeNet) | test error: 25.52% -> 24.45% | Improved |
| add CWN to plain GoogLeNet | plain · ImageNet-2012 val (top-1) | top-1 error: 30.78 -> 26.1 | Improved |
| add CWN to plain GoogLeNet | plain · ImageNet-2012 val (top-5) | top-5 error: 11.14 -> 8.35 | Improved |
| add centering (CWN) on top of batch normalization (CWN+BN vs WN+BN, NNN+BN) (ablation) | BN · SVHN permutation-invariant (MLP) | speed and test error | Improved |
Closest papers by meaning
- Weight Normalization: A Simple Reparameterization to Accelerate Training of Deep Neural Networks (NeurIPS 2016 · similarity 0.90)
- Optimization Theory for ReLU Neural Networks Trained with Normalization Layers (ICML 2020 · similarity 0.90)
- Weight Conditioning for Smooth Optimization of Neural Networks (ECCV 2024 · similarity 0.89)
- Norm matters: efficient and accurate normalization schemes in deep networks (NeurIPS 2018 · similarity 0.88)
- Neural Characteristic Activation Analysis and Geometric Parameterization for ReLU Networks (NeurIPS 2024 · similarity 0.85)
- Compelling ReLU Networks to Exhibit Exponentially Many Linear Regions at Initialization and During Training (ICML 2025 · similarity 0.85)
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.
