Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Few-Shot Learning With Global Class Representations
Aoxue Li, Tiange Luo, Tao Xiang, Weiran Huang, Liwei Wang
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 registration module and sample synthesis together (ablation) | global class representation baseline B · miniImageNet 5-way (ablation) | accuracy | Improved |
| add sample synthesis alone (ablation) | global representation baseline B · miniImageNet 5-way (ablation) | accuracy | No clear effect |
| add registration module alone (ablation) | global representation baseline B · miniImageNet 5-way (ablation) | accuracy | No clear effect |
| use soft weighted selection instead of argmax in registration (ablation) | registration module · miniImageNet | accuracy: 100% registration accuracy | No clear effect |
| learn global class representations with joint base-novel training instead of episodic prototypes | Prototypical Nets · miniImageNet 5-way 5-shot | accuracy: 68.20 -> 72.34 | Improved |
| global representation approach vs strongest baseline on 1-shot | MMN · miniImageNet 5-way 1-shot | accuracy: 53.37 vs 53.21 | No clear effect |
| global representation approach vs MG | MG · Omniglot 20-way 1-shot | accuracy: 97.64 -> 99.63 | Improved |
| global representation approach vs MG | MG · Omniglot 5-way 5-shot | accuracy: 99.86 -> 99.90 | No clear effect |
| use global class representations instead of episodic ones | PN · miniImageNet generalized | novel class accuracy: 0.52 -> 12.98 | Improved |
| use global class representations instead of episodic ones | RN · miniImageNet generalized | overall accuracy: 32.48 -> 39.14 | Improved |
| fix 100 seen class representations and add 20 new novel classes | Ours vs PN · miniImageNet + 20 new classes | novel class accuracy: 1.07 -> 12.60 | Improved |
| global representation approach vs MN | MN · miniImageNet generalized | novel class accuracy: 0.75 -> 12.98 | Improved |
Closest papers by meaning
- Few-Shot Learning With Embedded Class Models and Shot-Free Meta Training (ICCV 2019 · similarity 0.90)
- Few-Shot Learning With a Strong Teacher (TPAMI 2022 · similarity 0.89)
- Few-Shot Class-Incremental Learning via Training-Free Prototype Calibration (NeurIPS 2023 · similarity 0.88)
- Scaling Few-Shot Learning for the Open World (AAAI 2024 · similarity 0.88)
- Few-Shot Class-Incremental Learning via Class-Aware Bilateral Distillation (CVPR 2023 · similarity 0.88)
- Pseudo Informative Episode Construction for Few-Shot Class-Incremental Learning (AAAI 2025 · similarity 0.88)
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.
