Research map / Label-efficient and robust learning / Self-supervised and contrastive
Few-Shot Classification with Contrastive Learning
Zhanyuan Yang, Jinghua Wang, Yingying Zhu
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 all pre-training contrastive losses to CE (ablation) | FEAT baseline · miniImageNet 5-way 1-shot | accuracy: 66.58 -> 69.53 | Improved |
| add all pre-training contrastive losses to CE (ablation) | FEAT baseline · miniImageNet 5-way 5-shot | accuracy: 81.92 -> 84.33 | Improved |
| add cross-view episodic training (ablation) | L_CE+L_CL · miniImageNet 5-way 1-shot | accuracy: 69.53 -> 69.89 | Improved |
| add cross-view episodic training (ablation) | L_CE+L_CL · CIFAR-FS 5-way 1-shot | accuracy: 76.15 -> 77.02 | Improved |
| add distance-scaled contrastive loss L_info (ablation) | L_CE+L_CL · tieredImageNet 5-way 1-shot | accuracy: 71.83 -> 72.51 | Improved |
| add both CVET and L_info (ablation) | L_CE+L_CL · miniImageNet 5-way 1-shot | accuracy: 69.53 -> 70.19 | Improved |
| add local self-supervised loss only (ablation) | L_CE · miniImageNet 5-way 1-shot | accuracy: 66.58 -> 69.33 | Improved |
| add global supervised contrastive loss only (ablation) | L_CE · miniImageNet 5-way 1-shot | accuracy: 66.58 -> 67.88 | Improved |
| add vector-map module (ablation) | global contrastive losses · miniImageNet 5-way 1-shot | accuracy: 68.68 -> 69.29 | Improved |
| add map-map module (ablation) | global contrastive losses · miniImageNet 5-way 5-shot | accuracy: 84.17 -> 84.25 | Improved |
| use SimCLR-style instead of standard augmentation (ablation) | Ours · miniImageNet 5-way 1-shot | accuracy: 67.84 -> 70.19 | Improved |
| use AutoAugment instead of SimCLR-style augmentation (ablation) | Ours · miniImageNet 5-way 1-shot | accuracy: 70.19 -> 68.45 | Hurt |
Closest papers by meaning
- Pushing the Limit of Fine-Tuning for Few-Shot Learning: Where Feature Reusing Meets Cross-Scale Attention (AAAI 2024 · similarity 0.88)
- BECLR: Batch Enhanced Contrastive Few-Shot Learning (ICLR 2024 · similarity 0.88)
- Learning a Few-shot Embedding Model with Contrastive Learning (AAAI 2021 · similarity 0.88)
- A Simple Framework for Contrastive Learning of Visual Representations (ICML 2020 · similarity 0.88)
- ConFeSS: A Framework for Single Source Cross-Domain Few-Shot Learning (ICLR 2022 · similarity 0.88)
- Self-Supervised Contrastive Learning is Approximately Supervised Contrastive Learning (NeurIPS 2025 · similarity 0.87)
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.
