Research map / Architectures and efficiency / CNN design
Data Representations' Study of Latent Image Manifolds
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 |
|---|---|---|---|
| Train ResNet50 with mixup | ResNet50 · CIFAR10/CIFAR100 | data-manifold curvature | Helped |
| Train ResNet50 with manifold mixup | ResNet50 · CIFAR10/CIFAR100 | data-manifold curvature | No clear effect |
| Use different intrinsic-dimension estimation methods | ResNet50 · CIFAR10 | MAPC | No clear effect |
| Compare intrinsic-dimension profiles with MAPC profiles across layers | trained deep convolutional networks · CIFAR10/CIFAR100 and ImageNet | correlation between intrinsic dimension and MAPC | No clear effect |
| Use the normalized MAPC gap as a generalization indicator | trained models · CIFAR10/CIFAR100 and Tiny ImageNet | generalization ability | Helped |
| Evaluate MAPC, MAMC, MARC, and MASC instead of relying on one curvature metric | curvature analysis · Appendix A, Fig. 6 | curvature profile | No clear effect |
Closest papers by meaning
- Deep Networks and the Multiple Manifold Problem (ICLR 2021 · similarity 0.89)
- Curvature-Balanced Feature Manifold Learning for Long-Tailed Classification (CVPR 2023 · similarity 0.89)
- Empirical Study of the Topology and Geometry of Deep Networks (CVPR 2018 · similarity 0.89)
- Improving Neural Network Surface Processing with Principal Curvatures (NeurIPS 2024 · similarity 0.88)
- Building Transformation Layers for Riemannian Neural Networks (NeurIPS 2026 · similarity 0.87)
- Connecting Neural Models Latent Geometries with Relative Geodesic Representations (NeurIPS 2025 · similarity 0.86)
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.