Research map / Label-efficient and robust learning / Semi-supervised learning
Transporting Labels via Hierarchical Optimal Transport for Semi-Supervised Learning
Fariborz Taherkhani, Ali Dabouei, Sobhan Soleymani, Jeremy Dawson, Nasser M. Nasrabadi
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 |
|---|---|---|---|
| cluster with Wasserstein barycenters instead of k-means (ablation) | HrOT · CIFAR-10 (1000 labels) | test error (%): 15.78 -> 11.91 | Improved |
| cluster with Wasserstein barycenters instead of k-means (ablation) | HrOT · CIFAR-100 (4000 labels) | test error (%): 42.06 -> 38.98 | Improved |
| add hierarchical OT over clusters rather than direct OT (DOT) (ablation) | HrOT · CIFAR-10 (1000 labels) | test error (%): 17.97 -> 11.91 | Improved |
| add hierarchical OT over clusters rather than direct OT (DOT) (ablation) | HrOT · SVHN (1000 labels) | test error (%): 4.11 -> 3.06 | Improved |
| add hierarchical OT over clusters rather than direct OT (DOT) (ablation) | HrOT · Mini-ImageNet top-1 (4000 labels) | test error (%): 73.84 -> 67.66 | Improved |
| add center loss (ablation) | HrOT · CIFAR-10 (1000 labels) | test error (%): 13.65 -> 11.91 | Improved |
| add center loss (ablation) | HrOT · SVHN (1000 labels) | test error (%): 3.61 -> 3.06 | Improved |
| add center loss (ablation) | HrOT · CIFAR-100 (4000 labels) | test error (%): 40.66 -> 38.98 | Improved |
| use one-hot pseudo-labels instead of soft pseudo-labels (ablation) | HrOT · CIFAR-10 (1000 labels) | test error (%): 12.58 -> 11.91 | Improved |
| use one-hot pseudo-labels instead of soft pseudo-labels (ablation) | HrOT · CIFAR-100 (4000 labels) | test error (%): 40.02 -> 38.98 | Improved |
| use one-hot pseudo-labels instead of soft pseudo-labels (ablation) | HrOT · SVHN (1000 labels) | test error (%): 3.55 -> 3.06 | Improved |
| add HrOT pseudo-labels to supervised training (ablation) | 13-layer CNN · CIFAR-10 (4000 labels) | test error (%): 21.79 -> 7.74 | Improved |
Closest papers by meaning
- OTMatch: Improving Semi-Supervised Learning with Optimal Transport (ICML 2024 · similarity 0.89)
- Unsupervised Selective Labeling for More Effective Semi-Supervised Learning (ECCV 2022 · similarity 0.89)
- Realistic Evaluation of Deep Semi-Supervised Learning Algorithms (NeurIPS 2018 · similarity 0.88)
- All Labels Are Not Created Equal: Enhancing Semi-Supervision via Label Grouping and Co-Training (CVPR 2021 · similarity 0.88)
- Towards Realistic Semi-Supervised Learning (ECCV 2022 · similarity 0.88)
- InterLUDE: Interactions between Labeled and Unlabeled Data to Enhance Semi-Supervised Learning (ICML 2024 · similarity 0.87)
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.
