Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Prototype Completion for Few-Shot Learning
Baoquan Zhang, Xutao Li, Yunming Ye, Shanshan Feng
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 ProtoComNet prototype completion (ablation) | mean-based prototypes · miniImagenet 5-way 1-shot | accuracy: 61.22 -> 65.62 | Improved |
| add ProtoComNet prototype completion (ablation) | mean-based prototypes · miniImagenet 5-way 5-shot | accuracy: 78.72 -> 75.32 | Hurt |
| add PATNet to model unseen parts/attributes (ablation) | ProtoComNet · miniImagenet 5-way 1-shot | accuracy: 65.62 -> 66.52 | Improved |
| add PATNet to model unseen parts/attributes (ablation) | ProtoComNet · tieredImagenet 5-way 5-shot | accuracy: 80.78 -> 84.10 | Improved |
| add MeanFusion of mean and completed prototypes (ablation) | ProtoComNet + PATNet · miniImagenet 5-way 5-shot | accuracy: 75.68 -> 81.65 | Improved |
| replace MeanFusion with two-step GaussFusion (ablation) | ProtoComNet + PATNet · miniImagenet 5-way 1-shot | accuracy: 69.68 -> 76.05 | Improved |
| replace two-step estimation with improved EM in GaussFusion (ablation) | GaussFusion · miniImagenet 5-way 1-shot | accuracy: 76.05 -> 79.01 | Improved |
| replace two-step estimation with improved EM in GaussFusion (ablation) | GaussFusion · CUB-200-2011 5-way 1-shot | accuracy: 93.30 -> 93.78 | Improved |
| add GaussFusion under primitive-knowledge noise gamma=0.3 (ablation) | prototype completion · miniImagenet 5-way 1-shot | accuracy: 42.53 -> 77.24 | Improved |
| add GaussFusion instead of MeanFusion under noise gamma=0.3 (ablation) | prototype completion · tieredImagenet 5-way 1-shot | accuracy: 57.77 -> 81.51 | Improved |
| add ProtoComNet prototype completion (ablation) | mean-based prototypes · CUB-200-2011 5-way 5-shot | accuracy: 91.36 -> 89.19 | Hurt |
| use EM-based estimation vs improved EM (ablation) | GaussFusion · miniImagenet 5-way 1-shot (transductive) | accuracy: 75.35 -> 79.01 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- MetaBaseline — Meta-Baseline: Exploring Simple Meta-Learning for Few-Shot Learning (ICCV 2021) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Prototype Completion With Primitive Knowledge for Few-Shot Learning (CVPR 2021 · similarity 0.98)
- A Closer Look at Prototype Classifier for Few-shot Image Classification (NeurIPS 2022 · similarity 0.90)
- Pushing the Limit of Fine-Tuning for Few-Shot Learning: Where Feature Reusing Meets Cross-Scale Attention (AAAI 2024 · similarity 0.87)
- Fine-Grained Prototypes Distillation for Few-Shot Object Detection (AAAI 2024 · similarity 0.86)
- Mind the Gap Between Prototypes and Images in Cross-domain Finetuning (NeurIPS 2024 · similarity 0.86)
- Envisioning Class Entity Reasoning by Large Language Models for Few-shot Learning (AAAI 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.
