Research map / Architectures and efficiency / RNNs and normalization
Online Normalization for Training Neural Networks
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 Batch Norm with Online Normalization | ResNet-20 · CIFAR-10 | accuracy (%): 92.2 -> 92.3 | No clear effect |
| replace Batch Norm with Online Normalization | ResNet-20 · CIFAR-100 | loss: 1.14 -> 1.12 | Helped |
| replace Batch Norm with Online Normalization | ResNet-50 · ImageNet | accuracy (%): 76.4 -> 76.3 | No clear effect |
| replace Batch Norm with Online Normalization | ResNet-50 · ImageNet | loss: 0.97 -> 0.94 | Helped |
| replace Group Norm with Online Normalization | ResNet-20 · CIFAR-10 | accuracy (%): 90.3 -> 92.3 | Helped |
| replace Layer Norm with Online Normalization | ResNet-20 · CIFAR-100 | accuracy (%): 59.2 -> 68.6 | Helped |
| drop batching (Online Normalization vs Batch Norm batch 32) | U-Net 250^3 voxels · training memory | memory (GB): 195 -> 6 | Helped |
| replace other normalizers with Online Normalization | U-Net · synthetic shapes segmentation | Jaccard | Helped |
| replace other normalizers with Online Normalization | RNN and LSTM · PTB language modeling | perplexity | Helped |
Closest papers by meaning
- Normalizing the Normalizers: Comparing and Extending Network Normalization Schemes (ICLR 2017 · similarity 0.92)
- Towards Understanding Regularization in Batch Normalization (ICLR 2019 · similarity 0.90)
- Understanding Batch Normalization (NeurIPS 2018 · similarity 0.89)
- Four Things Everyone Should Know to Improve Batch Normalization (ICLR 2020 · similarity 0.89)
- Regularizing by the Variance of the Activations' Sample-Variances (NeurIPS 2018 · similarity 0.88)
- Towards Training Without Depth Limits: Batch Normalization Without Gradient Explosion (ICLR 2024 · similarity 0.87)
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.