Research map / Label-efficient and robust learning / Semi-supervised learning
A Closer Look at Embedding Propagation for Manifold Smoothing
Diego Velazquez, Pau Rodriguez, Josep M. Gonfaus, F. Xavier Roca, Jordi Gonzalez
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 embedding propagation (test time) | ResNet-18 · CIFAR10 PGD attack | accuracy under attack: 0.006 -> 11.70 | Improved |
| add embedding propagation (test time) | ResNet-18 · CIFAR10 FAB attack | accuracy under attack: 0.58 -> 5.64 | Improved |
| add embedding propagation (test time) | ResNet-18 · CIFAR10 FGSM attack | accuracy under attack: 17.24 -> 31.24 | Improved |
| add manifold mixup | ResNet-18 · CIFAR10 PGD attack | accuracy under attack: 0.006 -> 0.03 | No clear effect |
| add manifold mixup to EP | ResNet-18 · MNIST FAB attack | accuracy under attack: 14.20 -> 61.99 | Improved |
| add manifold mixup to EP | ResNet-18 · CIFAR10 PGD attack | accuracy under attack: 11.70 -> 8.79 | Hurt |
| add embedding propagation to MoCo pretraining | MoCo · STL-10 | accuracy: 85.28 -> 86.02 | Improved |
| add embedding propagation to MoCo pretraining | MoCo · CIFAR100 | accuracy: 74.68 -> 75.02 | Improved |
| use EP instead of manifold mixup in MoCo | MoCo · STL-10 | accuracy: 85.75 -> 86.02 | Improved |
| add embedding propagation (ablation) | EPNet WRN-28-10 · miniImagenet 1-shot | accuracy: 65.98 -> 70.74 | Improved |
| add embedding propagation (ablation) | EPNet WRN-28-10 · tieredImagenet 1-shot | accuracy: 74.04 -> 78.50 | Improved |
| add embedding propagation (ablation) | EPNet CONV-4 · miniImagenet 5-shot | accuracy: 72.57 -> 72.95 | Improved |
Closest papers by meaning
- Embedding Propagation: Smoother Manifold for Few-Shot Classification (ECCV 2020 · similarity 0.94)
- Manifold Mixup: Better Representations by Interpolating Hidden States (ICML 2019 · similarity 0.87)
- Hardness of Learning Neural Networks under the Manifold Hypothesis (NeurIPS 2024 · similarity 0.85)
- Diffusion Models and the Manifold Hypothesis: Log-Domain Smoothing is Geometry Adaptive (NeurIPS 2025 · similarity 0.85)
- Euclidean Embedding of Data Using Local Distances (NeurIPS 2026 · similarity 0.84)
- Intermediate Layer Classifiers for OOD generalization (ICLR 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.
