Research map / Label-efficient and robust learning / Noisy labels
Robust Meta-learning with Sampling Noise and Label Noise via Eigen-Reptile
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 Reptile's last-step update with an update along the main direction of historical task parameters | Reptile · Mini-Imagenet 5-way 1-shot | accuracy: 49.97 \pm 0.32\% -> 51.80 \pm 0.90\% | Helped |
| Replace Reptile's last-step update with an update along the main direction of historical task parameters | Reptile · Mini-Imagenet 5-way 5-shot | accuracy: 65.99 \pm 0.58\% -> 68.10\pm 0.50 \% | Helped |
| Use Eigen-Reptile instead of Reptile | Reptile · CIFAR-FS 5-way 1-shot | accuracy: 58.30 \pm 1.20 \% -> 61.90\pm1.40\% | Helped |
| Use Eigen-Reptile instead of Reptile | Reptile · CIFAR-FS 5-way 5-shot | accuracy: 75.45\pm 0.55\% -> 78.30 \pm 0.50\% | Helped |
| Use Eigen-Reptile instead of Reptile | Reptile · Mini-Imagenet 5-way 1-shot with symmetric label noise | accuracy: 47.64\% -> 47.87\% (p=0.0); 23.33\% -> 27.23\% (p=0.5) | Helped |
| Use Eigen-Reptile instead of Reptile | Reptile · Mini-Imagenet 5-way 1-shot with asymmetric label noise | accuracy: 42.03\% -> 42.29\% (p=0.5) | Helped |
| Add ISPL sample selection to Eigen-Reptile (ablation) | Eigen-Reptile · Mini-Imagenet 5-way 1-shot, symmetric label noise | accuracy: 27.23\% -> 28.68\% (p=0.5) | Helped |
| Add ISPL sample selection to Eigen-Reptile (ablation) | Eigen-Reptile · Mini-Imagenet 5-way 1-shot, asymmetric label noise | accuracy: 42.29\% -> 43.71\% (p=0.5) | Helped |
| Add ISPL sample selection to Eigen-Reptile (ablation) | Eigen-Reptile · Mini-Imagenet 5-way 1-shot without label noise | accuracy: 47.87\% -> 47.26\% | Hurt |
| Add ISPL sample selection to Reptile (ablation) | Reptile · Mini-Imagenet 5-way 1-shot with label noise | accuracy: 47.64\% -> 47.23\% (p=0.0); 23.33\% -> 21.83\% (symmetric p=0.5); 42.03\% -> 41.09\% (asymmetric p=0.5) | Mixed |
| Use the 64-filter version instead of the 32-filter version (ablation) | Eigen-Reptile · Mini-Imagenet 5-way 1-shot | accuracy: 51.80 \pm 0.90\% -> 53.25 \pm 0.45\% | Helped |
| Use the 64-filter version instead of the 32-filter version (ablation) | Eigen-Reptile · Mini-Imagenet 5-way 5-shot | accuracy: 68.10\pm 0.50 \% -> 69.85\pm 0.85 \% | Helped |
Closest papers by meaning
- Few-Shot Learning With Noisy Labels (CVPR 2022 · similarity 0.86)
- A Sample Complexity Separation between Non-Convex and Convex Meta-Learning (ICML 2020 · similarity 0.86)
- Dual-Level Curriculum Meta-Learning for Noisy Few-Shot Learning Tasks (AAAI 2024 · similarity 0.86)
- MetaDiff: Meta-Learning with Conditional Diffusion for Few-Shot Learning (AAAI 2024 · similarity 0.85)
- Any-Way Meta Learning (AAAI 2024 · similarity 0.85)
- Learning from Noisy Labels via Self-Taught On-the-Fly Meta Loss Rescaling (AAAI 2025 · similarity 0.85)
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.