Research map / Architectures and efficiency / RNNs and normalization
Filter Response Normalization Layer: Eliminating Batch Dependence in the Training of Deep Neural Networks
Check 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 BatchNorm with FRN + TLU | ResnetV2-50 · ImageNet validation, 32 images/GPU | Precision@1: 76.21 -> 77.21 | Helped |
| replace BatchNorm with FRN + TLU | InceptionV3 · ImageNet validation, 32 images/GPU | Precision@1: 78.24 -> 78.95 | Helped |
| use FRN instead of GroupNorm | ResnetV2-50 · ImageNet validation, 1 image/GPU | Precision@1: 75.93 -> 77.36 | Helped |
| replace ReLU with TLU while retaining BatchNorm (ablation) | ResnetV2-50 · ImageNet validation | Precision@1: 76.21 -> 76.03 | No clear effect |
| replace InstanceNorm with FRN while retaining ReLU (ablation) | ResnetV2-50 · ImageNet validation | Precision@1: 71.63 -> 75.24 | Helped |
| add TLU to FRN + ReLU (ablation) | ResnetV2-50 · ImageNet validation | Precision@1: 75.24 -> 77.21 | Helped |
| replace ReLU with TLU while retaining InstanceNorm (ablation) | ResnetV2-50 · ImageNet validation | Precision@1: 71.63 -> 71.72 | Helped |
| replace GroupNorm with GFRN while retaining ReLU (ablation) | ResnetV2-50 · ImageNet validation | Precision@1: 75.67 -> 75.93 | Helped |
| add TLU to GFRN + ReLU (ablation) | ResnetV2-50 · ImageNet validation | Precision@1: 75.93 -> 76.44 | Helped |
| replace ReLU with TLU while retaining GroupNorm (ablation) | ResnetV2-50 · ImageNet validation | Precision@1: 75.67 -> 76.59 | Helped |
| replace LayerNorm with LFRN while retaining ReLU (ablation) | ResnetV2-50 · ImageNet validation | Precision@1: 72.75 -> 75.03 | Helped |
| add TLU to LFRN + ReLU (ablation) | ResnetV2-50 · ImageNet validation | Precision@1: 75.03 -> 76.17 | Helped |
Closest papers by meaning
- Understanding Batch Normalization (NeurIPS 2018 · similarity 0.91)
- Four Things Everyone Should Know to Improve Batch Normalization (ICLR 2020 · similarity 0.91)
- Towards Understanding Regularization in Batch Normalization (ICLR 2019 · similarity 0.91)
- Normalizing the Normalizers: Comparing and Extending Network Normalization Schemes (ICLR 2017 · similarity 0.90)
- Kalman Normalization: Normalizing Internal Representations Across Network Layers (NeurIPS 2018 · similarity 0.90)
- Batch normalization is sufficient for universal function approximation in CNNs (ICLR 2024 · similarity 0.90)
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.