Research map / Architectures and efficiency / RNNs and normalization
Decorrelated Batch Normalization
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 DBN in the VGG-A configuration | VGG-A · CIFAR-10 | absolute test error: 0.61% | Helped |
| Replace BN with DBN while using Adam | VGG-A · CIFAR-10 | absolute test error: 1.34% | Helped |
| Replace BN with DBN and use ELU instead of ReLU | VGG-A · CIFAR-10 | absolute test error: 1.44% | Helped |
| Replace BN with DBN after the nonlinearity | VGG-A · CIFAR-10 | absolute test error: 0.38% | Helped |
| Use PCA whitening instead of ZCA whitening (ablation) | DBN · MNIST, 2-layer and 4-layer MLP | training performance | Hurt |
| Use ZCA whitening instead of PCA whitening (ablation) | DBN · MNIST, 2-layer and 4-layer MLP | training performance | Helped |
| Control whitening with an intermediate group size (such as G16) instead of full whitening or BN-equivalent G1 (ablation) | DBN · PIE | training loss convergence | Helped |
| Add adjustable scale to DBN-L1 (ablation) | Residual networks · CIFAR-10, Res-20/32/44/56 | test error (%): 7.94 -> 7.77; 7.28 -> 6.94; 6.87 -> 6.83; 6.63 -> 6.49 | Helped |
Closest papers by meaning
- Iterative Normalization: Beyond Standardization Towards Efficient Whitening (CVPR 2019 · similarity 0.93)
- Stochastic Whitening Batch Normalization (CVPR 2021 · similarity 0.92)
- Understanding Batch Normalization (NeurIPS 2018 · similarity 0.92)
- Four Things Everyone Should Know to Improve Batch Normalization (ICLR 2020 · similarity 0.91)
- Towards Stabilizing Batch Statistics in Backward Propagation of Batch Normalization (ICLR 2020 · similarity 0.90)
- Reducing Divergence in Batch Normalization for Domain Adaptation (AAAI 2025 · 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.