Research map / Architectures and efficiency / CNN design
Exemplar Normalization for Learning Deep Representation
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 |
|---|---|---|---|
| Use sample-dependent exemplar normalization instead of switchable normalization | ResNet50 · ImageNet val | top-1 accuracy: 76.9 -> 78.1 | Helped |
| Use sample-dependent exemplar normalization instead of switchable normalization | ResNet50 · Webvision val | top-1 accuracy: 72.8 -> 73.5 | Helped |
| set pi to 1 (ablation) | EN-ResNet50 · ImageNet val | top-1 accuracy: 78.1 -> 77.1 | Hurt |
| set pi to 10 (ablation) | EN-ResNet50 · ImageNet val | top-1 accuracy: 78.1 -> 77.5 | Hurt |
| set pi to 20 (ablation) | EN-ResNet50 · ImageNet val | top-1 accuracy: 78.1 -> 77.8 | Hurt |
| set pi to 50 (ablation) | EN-ResNet50 · ImageNet val | top-1 accuracy: 78.1 -> 78.1 | No clear effect |
| set pi to 100 (ablation) | EN-ResNet50 · ImageNet val | top-1 accuracy: 78.1 -> 78.0 | Hurt |
| set r to 2 (ablation) | EN-ResNet50 · ImageNet val | top-1 accuracy: 78.1 -> 77.7 | Hurt |
| set r to 4 (ablation) | EN-ResNet50 · ImageNet val | top-1 accuracy: 78.1 -> 77.9 | Hurt |
| set r to 16 (ablation) | EN-ResNet50 · ImageNet val | top-1 accuracy: 78.1 -> 77.9 | Hurt |
| set r to 32 (ablation) | EN-ResNet50 · ImageNet val | top-1 accuracy: 78.1 -> 78.1 | No clear effect |
| set r to 64 (ablation) | EN-ResNet50 · ImageNet val | top-1 accuracy: 78.1 -> 77.7 | Hurt |
Closest papers by meaning
- Differentiable Learning-to-Normalize via Switchable Normalization (ICLR 2019 · similarity 0.89)
- Normalizing the Normalizers: Comparing and Extending Network Normalization Schemes (ICLR 2017 · similarity 0.89)
- Filter Response Normalization Layer: Eliminating Batch Dependence in the Training of Deep Neural Networks (CVPR 2020 · similarity 0.88)
- Differentiable Dynamic Normalization for Learning Deep Representation (ICML 2019 · similarity 0.88)
- On the Nonlinearity of Layer Normalization (ICML 2024 · similarity 0.85)
- Normalize Filters! Classical Wisdom for Deep Vision (NeurIPS 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.