Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Learning a Few-shot Embedding Model with 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 |
|---|---|---|---|
| Add supervised infoNCE to the baseline embedding training (ablation) | ResNet12 few-shot embedding · miniImageNet 5-way 1-shot | accuracy: 61.69 -> 64.23 | Helped |
| Add supervised infoNCE to the baseline embedding training (ablation) | ResNet12 few-shot embedding · miniImageNet 5-way 5-shot | accuracy: 78.31 -> 79.17 | Helped |
| Add patch-wise hard-sample training (ablation) | ResNet12 with infoNCE · miniImageNet 5-way 1-shot | accuracy: 64.23 -> 66.80 | Helped |
| Add patch-wise hard-sample training (ablation) | ResNet12 with infoNCE · miniImageNet 5-way 5-shot | accuracy: 79.17 -> 81.35 | Helped |
| Add PatchMix (ablation) | ResNet12 with infoNCE and hard-sample training · miniImageNet 5-way 1-shot | accuracy: 66.80 -> 67.67 | Helped |
| Add PatchMix (ablation) | ResNet12 with infoNCE and hard-sample training · miniImageNet 5-way 5-shot | accuracy: 81.35 -> 82.44 | Helped |
| Use a 6×6 or 11×11 patch grid rather than a 1×1 grid (ablation) | infoPatch · miniImageNet 5-way 1-shot | accuracy: 64.23 -> 66.19; 64.23 -> 66.80 | Helped |
| Use a 6×6 or 11×11 patch grid rather than a 1×1 grid (ablation) | infoPatch · miniImageNet 5-way 5-shot | accuracy: 79.17 -> 81.27; 79.17 -> 81.35 | Helped |
| Use share mix instead of independent mix (ablation) | infoPatch · miniImageNet 5-way 1-shot | accuracy: 67.67 -> 67.53 | Hurt |
| Use share mix instead of independent mix (ablation) | infoPatch · miniImageNet 5-way 5-shot | accuracy: 82.44 -> 81.94 | Hurt |
| Use exchange mix instead of independent mix (ablation) | infoPatch · miniImageNet 5-way 1-shot | accuracy: 67.67 -> 67.48 | Hurt |
| Use exchange mix instead of independent mix (ablation) | infoPatch · miniImageNet 5-way 5-shot | accuracy: 82.44 -> 82.06 | Hurt |
Closest papers by meaning
- BECLR: Batch Enhanced Contrastive Few-Shot Learning (ICLR 2024 · similarity 0.88)
- Few-Shot Learning via Embedding Adaptation With Set-to-Set Functions (CVPR 2020 · similarity 0.88)
- Pushing the Limit of Fine-Tuning for Few-Shot Learning: Where Feature Reusing Meets Cross-Scale Attention (AAAI 2024 · similarity 0.87)
- ConFeSS: A Framework for Single Source Cross-Domain Few-Shot Learning (ICLR 2022 · similarity 0.87)
- Few-Shot Recognition via Stage-Wise Retrieval-Augmented Finetuning (CVPR 2025 · similarity 0.85)
- Boosting Few-Shot Learning via Attentive Feature Regularization (AAAI 2024 · 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.