Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Unsupervised Meta-learning via Few-shot Pseudo-supervised Contrastive Learning
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 |
|---|---|---|---|
| Use PsCo rather than Meta-SVEBM | unsupervised few-shot classification · miniImageNet 5-way 5-shot | accuracy: 58.03 -> 63.26 | Helped |
| Use PsCo rather than Meta-SVEBM | unsupervised few-shot classification · Omniglot and miniImageNet standard benchmarks | accuracy | Helped |
| Add momentum network to PsCo (ablation) | PsCo · Omniglot 5-way 1-shot | accuracy: 90.32 -> 96.37 | Helped |
| Add predictor to PsCo (ablation) | PsCo · Omniglot 5-way 1-shot | accuracy: 90.21 -> 96.37 | Helped |
| Add Sinkhorn-Knopp assignment (ablation) | PsCo · Omniglot 5-way 1-shot | accuracy: 95.81 -> 96.37 | Helped |
| Add top-K support sampling (ablation) | PsCo · Omniglot 5-way 1-shot | accuracy: 94.95 -> 96.37 | Helped |
| Add MoCo objective (ablation) | PsCo · Omniglot 5-way 1-shot | accuracy: 93.16 -> 96.37 | Helped |
| Use weak rather than strong augmentation for momentum representations (ablation) | PsCo · miniImageNet 5-way 5-shot | accuracy: 60.04 -> 63.26 | Helped |
| Use weak rather than strong augmentation for query and momentum representations (ablation) | PsCo · miniImageNet few-shot tasks | accuracy: 63.26 -> 55.05 | Hurt |
| Set meta-training shot count K to 4 rather than 1 (ablation) | PsCo · miniImageNet 5-way 5-shot | accuracy: 61.84 -> 63.26 | Helped |
| Adapt projection and prediction heads using support samples (ablation) | PsCo · CropDiseases 5-way 5-shot | accuracy: 79.75 -> 88.24 | Helped |
| Adapt projection and prediction heads using support samples (ablation) | PsCo · miniImageNet 5-way 5-shot | accuracy: 63.26 -> 63.30 | No clear effect |
Closest papers by meaning
- Unsupervised Learning via Meta-Learning (ICLR 2019 · similarity 0.90)
- Learning to Self-Train for Semi-Supervised Few-Shot Classification (NeurIPS 2019 · similarity 0.90)
- Unsupervised Meta-Learning through Latent-Space Interpolation in Generative Models (ICLR 2021 · similarity 0.89)
- BECLR: Batch Enhanced Contrastive Few-Shot Learning (ICLR 2024 · similarity 0.87)
- Unsupervised Meta-Learning via In-Context Learning (ICLR 2025 · similarity 0.87)
- Self-Supervised Contrastive Learning is Approximately Supervised Contrastive Learning (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.