Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
How to Trust Unlabeled Data? Instance Credibility Inference for Few-Shot Learning
Yikai Wang, Li Zhang, Yuan Yao, Yanwei Fu
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 |
|---|---|---|---|
| use ICI credibility selection instead of random selection (ablation) | ICI-R · miniImageNet 1-shot (transductive) | accuracy: 67.54 -> 71.19 | Improved |
| use ICI selection instead of nearest-neighbor selection (ablation) | ICI-R · miniImageNet 1-shot (transductive) | accuracy: 69.80 -> 71.19 | Improved |
| use ICI selection instead of classifier confidence (ablation) | ICI-R · miniImageNet 1-shot (transductive) | accuracy: 70.57 -> 71.19 | Improved |
| use ICI (x-axis) selection instead of coefficient norm (y-axis) (ablation) | ICI-R · miniImageNet 5-shot (semi-supervised) | accuracy: 81.49 -> 82.32 | Improved |
| select samples iteratively (2 iterations of 3) rather than once (ablation) | ICI · miniImageNet | accuracy: ICI(3) with 6 images beats ICI(8) with 8 images | Improved |
| use logistic regression instead of kNN as the classifier (ablation) | ICI-R · miniImageNet 5-shot (transductive) | accuracy: 79.88 -> 83.27 | Improved |
| use logistic regression instead of SVM as the classifier (ablation) | ICI-R · miniImageNet 5-shot (transductive) | accuracy: 82.76 -> 83.27 | Improved |
| reduce features to d=5 with LLE instead of no reduction (d=512) (ablation) | ICI-R · miniImageNet 1-shot | accuracy: 67.08 -> 72.39 | Improved |
| use LLE instead of MDS for dimension reduction (ablation) | ICI-R · miniImageNet 1-shot | accuracy: 68.05 -> 72.39 | Improved |
| use LLE instead of spectral embedding (ablation) | ICI-R · miniImageNet 1-shot | accuracy: 72.43 -> 72.39 | No clear effect |
| add more unlabeled data (30/50 instead of 15/15) | ICI-R · miniImageNet 1-shot (semi-supervised) | accuracy: 72.32 -> 73.12 | Improved |
| apply ICI on CAN features | ICI-R · miniImageNet 1-shot | accuracy: 67.19 -> 70.53 | Improved |
Closest papers by meaning
- Instance Credibility Inference for Few-Shot Learning (CVPR 2020 · similarity 0.96)
- Worst Case Matters for Few-Shot Recognition (ECCV 2022 · similarity 0.88)
- Boosting Few-Shot Visual Learning With Self-Supervision (ICCV 2019 · similarity 0.88)
- Negatives Make a Positive: An Embarrassingly Simple Approach to Semi-Supervised Few-Shot Learning (TPAMI 2023 · similarity 0.88)
- Label, Verify, Correct: A Simple Few Shot Object Detection Method (CVPR 2022 · similarity 0.87)
- Few-Shot Recognition via Stage-Wise Retrieval-Augmented Finetuning (CVPR 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.
