Research map / Label-efficient and robust learning / Semi-supervised learning
Shoestring: Graph-Based Semi-Supervised Classification With Severely Limited Labeled Data
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 a cosine-similarity metric-learning network to TPN | TPN · miniImageNet 5-way 1-shot | accuracy: 52.78 -> 55.35 | Helped |
| Add a cosine-similarity metric-learning network to TPN | TPN · miniImageNet 5-way 5-shot | accuracy: 66.42 -> 70.01 | Helped |
| Add a cosine-similarity metric-learning network to TPN | TPN · tieredImageNet 5-way 1-shot | accuracy: 55.74 -> 56.79 | Helped |
| Add a cosine-similarity metric-learning network to TPN | TPN · tieredImageNet 5-way 5-shot | accuracy: 71.01 -> 73.71 | Helped |
| Add metric learning with class centroids and cosine similarity | GLP(AR) · Cora, 1 label per class | accuracy: 37.7(4.0) -> 69.8(3.7) | Helped |
| Add metric learning with class centroids and cosine similarity | GLP(AR) · CiteSeer, 1 label per class | accuracy: 37.4(19) -> 58.1(17) | Helped |
| Add metric learning with class centroids and cosine similarity | GLP(AR) · PubMed, 1 label per class | accuracy: 55.8(9.1) -> 65.2(7.7) | Helped |
| Add metric learning with class centroids and cosine similarity | IGCN(AR) · Large Cora, 1 label per class | accuracy: 56.0(8.3) -> 56.0(8.3) | Helped |
| Use L2 distance instead of cosine similarity in Shoestring (ablation) | Shoestring with graph-based citation-network models · various citation networks and label rates | accuracy | Mixed |
| Use L1 distance in the metric-learning network (ablation) | Shoestring with graph-based citation-network models · citation networks | accuracy | Mixed |
| Add Shoestring metric learning to graph-based methods | citation-network classification · 20 labels per class, Large Cora | accuracy: 68.8 -> 70.5 | Helped |
Closest papers by meaning
- Label Efficient Semi-Supervised Learning via Graph Filtering (CVPR 2019 · similarity 0.90)
- SimPLE: Similar Pseudo Label Exploitation for Semi-Supervised Classification (CVPR 2021 · similarity 0.88)
- Matrix Completion for Graph-Based Deep Semi-Supervised Learning (AAAI 2019 · similarity 0.87)
- Semi-Supervised Classification with Graph Convolutional Networks (ICLR 2017 · similarity 0.87)
- Data driven semi-supervised learning (NeurIPS 2021 · similarity 0.87)
- 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.