Research map / Image generation / Autoregressive and latent generation
On Feature Normalization and Data Augmentation
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 |
|---|---|---|---|
| Add moment exchange with paired label interpolation | ResNet-110 · CIFAR-100 | error (%): 26.28$\pm$0.10 -> 25.47$\pm$0.09 | Helped |
| Add MoEx to CutMix and ShakeDrop | PyramidNet-200 · CIFAR-100 | Top-1 error (%): 13.81 -> 13.47 | Helped |
| Add moment exchange | ResNet-50 · ImageNet, 300 epochs | test error (%): 23.1 -> 21.9 | Helped |
| Add MoEx to CutMix | ResNet-50 · ImageNet, 300 epochs | Top-1 error (%): 21.3 -> 20.9 | Helped |
| Pretrain the detector backbone with MoEx | Faster R-CNN with ResNet-50 C4 · PASCAL VOC 2007 test | APCOCO: 53.5 -> 54.6 | Helped |
| Add moment exchange | PointNet++ · ModelNet10 | classification error (%): 6.02$\pm$0.10 -> 5.25$\pm$0.18 | Helped |
| Disable label interpolation (lambda=1) (ablation) | MoEx on ResNet-110 · CIFAR-100 | test error (%): 25.5$\pm$0.09 -> 26.3$\pm$0.02 | Hurt |
| Replace paired label interpolation with label smoothing (ablation) | MoEx on ResNet-110 · CIFAR-100 | test error (%): 25.5$\pm$0.09 -> 25.8$\pm$0.09 | Hurt |
| Use LayerNorm moments instead of PONO moments (ablation) | MoEx on ResNet-110 · CIFAR-100 | test error (%): 25.5$\pm$0.09 -> 25.6$\pm$0.02 | No clear effect |
| Exchange only the first moment (ablation) | MoEx on ResNet-110 · CIFAR-100 | test error (%): 25.5$\pm$0.09 -> 25.9$\pm$0.06 | Hurt |
| Exchange only the second moment (ablation) | MoEx on ResNet-110 · CIFAR-100 | test error (%): 25.5$\pm$0.09 -> 26.0$\pm$0.13 | Hurt |
| Set exchange probability to 1 rather than 0.5 (ablation) | MoEx on ResNet-50 · ImageNet, 300 epochs | Top-1 error (%): 22.4 -> 21.9 | Helped |
Closest papers by meaning
- Soft Augmentation for Image Classification (CVPR 2023 · similarity 0.86)
- Learning Implicit Generative Models with the Method of Learned Moments (ICML 2018 · similarity 0.85)
- MCA: Moment Channel Attention Networks (AAAI 2024 · similarity 0.84)
- Characterizing the Generalization Error of Random Feature Regression with Arbitrary Data-Augmentation (NeurIPS 2026 · similarity 0.84)
- Moment- and Power-Spectrum-Based Gaussianity Regularization for Text-to-Image Models (NeurIPS 2025 · similarity 0.83)
- Exact Gaussian Moment Matching for Residual Networks: a Second-Order Method (NeurIPS 2026 · similarity 0.83)
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.