Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Meta-Learning for Semi-Supervised Few-Shot Classification
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 |
|---|---|---|---|
| Refine prototypes with unlabeled examples using soft k-Means | Prototypical Network · miniImageNet, no distractors | accuracy: 43.61 $\pm$ 0.27 -> 50.09 $\pm$ 0.45; 59.08 $\pm$ 0.22 -> 64.59 $\pm$ 0.28 | Helped |
| Train the embedding with semi-supervised refinement, rather than applying refinement only at test time | Soft k-Means · miniImageNet, no distractors | accuracy: 48.98 $\pm$ 0.34 -> 50.09 $\pm$ 0.45; 63.77 $\pm$ 0.20 -> 64.59 $\pm$ 0.28 | Helped |
| Add a distractor cluster to soft k-Means (ablation) | Soft k-Means · Omniglot 1-shot with distractors | accuracy: 95.01 $\pm$ 0.09 -> 97.17 $\pm$ 0.04 | Helped |
| Add a distractor cluster to soft k-Means (ablation) | Soft k-Means · miniImageNet with distractors | accuracy: 48.70 $\pm$ 0.32 -> 48.86 $\pm$ 0.32; 63.55 $\pm$ 0.28 -> 61.27 $\pm$ 0.24 | Mixed |
| Add a distractor cluster to soft k-Means (ablation) | Soft k-Means · tieredImageNet with distractors | accuracy: 49.88 $\pm$ 0.52 -> 51.36 $\pm$ 0.31; 68.32 $\pm$ 0.22 -> 67.56 $\pm$ 0.10 | Mixed |
| Mask distant unlabeled examples during refinement instead of assigning all examples to class clusters (ablation) | Soft k-Means · Omniglot 1-shot with distractors | accuracy: 95.01 $\pm$ 0.09 -> 97.30 $\pm$ 0.08 | Helped |
| Mask distant unlabeled examples during refinement (ablation) | Soft k-Means · miniImageNet with distractors | accuracy: 48.70 $\pm$ 0.32 -> 49.04 $\pm$ 0.31; 63.55 $\pm$ 0.28 -> 62.96 $\pm$ 0.14 | Mixed |
| Mask distant unlabeled examples during refinement (ablation) | Soft k-Means · tieredImageNet with distractors | accuracy: 49.88 $\pm$ 0.52 -> 51.38 $\pm$ 0.38; 68.32 $\pm$ 0.22 -> 69.08 $\pm$ 0.25 | Helped |
| Repeat prototype refinement for multiple iterations instead of one (ablation) | Soft k-Means · Experiments across datasets | classification performance | No clear effect |
| Increase the test-time unlabeled set per class from 0 to 25 (ablation) | Semi-supervised models · tieredImageNet | accuracy | Helped |
Closest papers by meaning
- Learning to Self-Train for Semi-Supervised Few-Shot Classification (NeurIPS 2019 · similarity 0.93)
- A Theoretical Analysis of the Number of Shots in Few-Shot Learning (ICLR 2020 · similarity 0.91)
- Meta-Dataset: A Dataset of Datasets for Learning to Learn from Few Examples (ICLR 2020 · similarity 0.91)
- A Dual Attention Network with Semantic Embedding for Few-Shot Learning (AAAI 2019 · similarity 0.90)
- The Role of Global Labels in Few-Shot Classification and How to Infer Them (NeurIPS 2021 · similarity 0.90)
- Pushing the Limit of Fine-Tuning for Few-Shot Learning: Where Feature Reusing Meets Cross-Scale Attention (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.