Research map / Label-efficient and robust learning / Semi-supervised learning
PTN: A Poisson Transfer Network for Semi-supervised Few-shot Learning
Huaxi Huang, Junjie Zhang, Jian Zhang, Qiang Wu, Chang Xu
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 |
|---|---|---|---|
| replace Label Propagation with PoissonMBO (ablation) | graph-based few-shot inference · miniImageNet 5-way 1-shot | accuracy: 74.04$\pm$1.00 -> 79.67$\pm$1.02 | Improved |
| add query feature calibration before PoissonMBO inference (ablation) | PoissonMBO few-shot classifier · miniImageNet 5-way 1-shot | accuracy: 79.67$\pm$1.02 -> 80.00$\pm$0.83 | Improved |
| add unsupervised embedding transfer to Label Propagation (ablation) | Label Propagation · miniImageNet 5-way 1-shot | accuracy: 74.04$\pm$1.00 -> 75.65$\pm$1.06 | Improved |
| add unsupervised embedding transfer to Label Propagation (ablation) | Label Propagation · miniImageNet 5-way 5-shot | accuracy: 82.60$\pm$0.68 -> 84.46$\pm$0.68 | Improved |
| add unsupervised embedding transfer to PoissonMBO (ablation) | PoissonMBO · miniImageNet 5-way 1-shot | accuracy: 79.67$\pm$1.02 -> 80.73$\pm$1.11 | Improved |
| add unsupervised embedding transfer to PoissonMBO (ablation) | PoissonMBO · miniImageNet 5-way 5-shot | accuracy: 86.30$\pm$0.65 -> 87.41$\pm$0.63 | Improved |
| increase extra unlabeled samples from 0 to 200 (ablation) | PTN · miniImageNet 5-way 1-shot | accuracy: 77.01$\pm$0.94 -> 82.66$\pm$0.97 | Improved |
| increase extra unlabeled samples from 0 to 200 (ablation) | PTN · miniImageNet 5-way 5-shot | accuracy: 85.32$\pm$0.68 -> 88.43$\pm$0.76 | Improved |
| use unlabeled query samples for embedding transfer when Num_U is 0 (ablation) | PTN versus PTN* · miniImageNet 5-way 1-shot | accuracy: 76.20$\pm$0.82 -> 77.01$\pm$0.94 | Improved |
| add distractor classes among unlabeled samples (ablation) | PTN · miniImageNet 5-way 1-shot | accuracy: 82.66$\pm$0.97 -> 81.92$\pm$1.02 | Hurt |
| add distractor classes among unlabeled samples (ablation) | PTN · tieredImageNet 5-way 5-shot | accuracy: 89.14$\pm$0.71 -> 88.06$\pm$0.62 | Hurt |
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.
- TransMatch — TransMatch: A Transfer-Learning Scheme for Semi-Supervised Few-Shot Learning (CVPR 2020) · 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
- Poisson Learning: Graph Based Semi-Supervised Learning At Very Low Label Rates (ICML 2020 · similarity 0.90)
- Contrastive Graph Poisson Networks: Semi-Supervised Learning with Extremely Limited Labels (NeurIPS 2021 · similarity 0.89)
- Variance-enlarged Poisson Learning for Graph-based Semi-Supervised Learning with Extremely Sparse Labeled Data (ICLR 2024 · similarity 0.88)
- Negatives Make a Positive: An Embarrassingly Simple Approach to Semi-Supervised Few-Shot Learning (TPAMI 2023 · similarity 0.88)
- An Embarrassingly Simple Approach to Semi-Supervised Few-Shot Learning (NeurIPS 2022 · similarity 0.88)
- InterLUDE: Interactions between Labeled and Unlabeled Data to Enhance Semi-Supervised Learning (ICML 2024 · 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.
