Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Robust Meta-Representation Learning via Global Label Inference and Classification
Ruohan Wang, Isak Falk, Massimiliano Pontil, Carlo Ciliberto
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 rotation-augmented pre-training (4x classes) (ablation) | MeLa pre-training · miniImageNet 1-shot | accuracy (%): 62.0 -> 64.5 | Improved |
| add rotation-augmented pre-training (4x classes) (ablation) | MeLa pre-training · miniImageNet 5-shot | accuracy (%): 79.6 -> 81.5 | Improved |
| add rotation-augmented pre-training (4x classes) (ablation) | MeLa pre-training · H-Aircraft 1-shot | accuracy (%): 79.2 -> 84.8 | Improved |
| add rotation-augmented pre-training (4x classes) (ablation) | MeLa pre-training · tieredImageNet 1-shot | accuracy (%): 69.1 -> 69.5 | No clear effect |
| add rotation-augmented pre-training (4x classes) (ablation) | MeLa pre-training · tiered-780 1-shot | accuracy (%): 78.0 -> 78.2 | No clear effect |
| use 640 classes with 60 samples per class instead of 64 classes (same dataset size) (ablation) | standard pre-training · miniImageNet vs mini-60 1-shot | accuracy (%): 62.0 -> 63.9 | Improved |
| add local-constraint-aware clustering instead of K-means (ablation) | MeLa pre-training · tieredImageNet 5-shot | accuracy (%): 78.8 -> 84.3 | Improved |
| add local-constraint-aware clustering instead of K-means (ablation) | MeLa pre-training · miniImageNet 5-shot | accuracy (%): 76.9 -> 81.5 | Improved |
| use no-replacement task sampling instead of with-replacement (ablation) | MeLa · miniImageNet 1-shot | accuracy (%): 65.8 -> 65.8 | No clear effect |
| raise pruning threshold q from 3 to 5 (ablation) | MeLa pre-training · miniImageNet 5-shot | accuracy (%): 81.5 -> 80.9 | No clear effect |
| add meta fine-tuning with residual connections to pre-training (ablation) | MeLa · miniImageNet 1-shot | accuracy (%): 64.5 -> 65.8 | Improved |
| add meta fine-tuning with residual connections to pre-training (ablation) | MeLa · tieredImageNet 5-shot | accuracy (%): 84.3 -> 85.9 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Meta-Baseline — Meta-Baseline: Exploring Simple Meta-Learning for Few-Shot Learning (ICCV 2021) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- The Role of Global Labels in Few-Shot Classification and How to Infer Them (NeurIPS 2021 · similarity 0.96)
- Revisiting Unsupervised Meta-Learning via the Characteristics of Few-Shot Tasks (TPAMI 2022 · similarity 0.91)
- Unraveling Meta-Learning: Understanding Feature Representations for Few-Shot Tasks (ICML 2020 · similarity 0.90)
- Is Meta-Learning Out? Rethinking Unsupervised Few-Shot Classification with Limited Entropy (ICCV 2025 · similarity 0.90)
- Unsupervised Meta-Learning via In-Context Learning (ICLR 2025 · similarity 0.89)
- Any-Way Meta Learning (AAAI 2024 · similarity 0.89)
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.
