Research map / Label-efficient and robust learning / Domain adaptation and generalization
Transferable Normalization: Towards Improving Transferability of Deep 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 BN with TransNorm | DANN · Office-31 avg | accuracy: 82.2 -> 86.0 | Helped |
| replace BN with TransNorm | CDAN · Office-31 avg | accuracy: 87.7 -> 89.3 | Helped |
| replace BN with TransNorm | CDAN · Office-Home avg | accuracy: 65.8 -> 67.6 | Helped |
| replace BN with TransNorm | CDAN · ImageCLEF-DA avg | accuracy: 87.7 -> 88.5 | Helped |
| replace BN with TransNorm | MDD · Office-31 avg | accuracy: 88.8 -> 89.5 | Helped |
| replace BN with TransNorm | CDAN (DTN) · SVHN->MNIST | accuracy: 89.2 -> 94.7 | Helped |
| replace BN with TransNorm | DANN · VisDA-2017 | accuracy: 63.7 -> 66.3 | Helped |
| replace BN with AdaBN | CDAN · Office-31 avg | accuracy: 87.7 -> 86.8 | Hurt |
| use learned channel transferability alpha instead of alpha=1 (ablation) | CDAN+TransNorm · Office-31 avg | accuracy: 88.5 -> 89.3 | Helped |
| use variance-normalized mean distance instead of Wasserstein (ablation) | CDAN+TransNorm · Office-31 avg | accuracy: 88.4 -> 89.3 | Helped |
| use variance-normalized mean distance instead of mean-only (ablation) | CDAN+TransNorm · Office-31 avg | accuracy: 88.9 -> 89.3 | Helped |
| use Student-t instead of softmax for channel probability (ablation) | CDAN+TransNorm · Office-31 avg | accuracy: 87.9 -> 89.3 | Helped |
Closest papers by meaning
- Transferable Adversarial Training: A General Approach to Adapting Deep Classifiers (ICML 2019 · similarity 0.87)
- Deep Transfer Learning with Joint Adaptation Networks (ICML 2017 · similarity 0.87)
- Reducing Divergence in Batch Normalization for Domain Adaptation (AAAI 2025 · similarity 0.87)
- Interpretations of Domain Adaptations via Layer Variational Analysis (ICLR 2023 · similarity 0.87)
- HybridNorm: Towards Stable and Efficient Transformer Training via Hybrid Normalization (NeurIPS 2025 · similarity 0.85)
- On Transferring Transferability: Towards a Theory for Size Generalization (NeurIPS 2025 · similarity 0.84)
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.