Research map / Label-efficient and robust learning / Semi-supervised learning
Transductive Semi-Supervised Deep Learning using Min-Max Features
Weiwei Shi, Yihong Gong, Chris Ding, Zhiheng MaXiaoyu Tao, Nanning Zheng
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 confidence levels on unlabeled samples (ablation) | TSSDL · CIFAR10 (1000 labels) | top-1 error rate (%): 50.38 -> 21.13 | Improved |
| add confidence levels on unlabeled samples (ablation) | TSSDL · CIFAR10 (4000 labels) | top-1 error rate (%): 18.10 -> 10.90 | Improved |
| add confidence levels on unlabeled samples (ablation) | TSSDL · SVHN (250 labels) | top-1 error rate (%): 46.74 -> 5.02 | Improved |
| add confidence levels on unlabeled samples (ablation) | TSSDL · SVHN (1000 labels) | top-1 error rate (%): 10.10 -> 3.80 | Improved |
| add MMF regularization to the plain transductive DCNN (ablation) | TDCNN · CIFAR10 (4000 labels) | top-1 error rate (%): 16.17 -> 13.11 | Improved |
| add RF regularization to the plain transductive DCNN (ablation) | TDCNN · CIFAR10 (4000 labels) | top-1 error rate (%): 16.17 -> 13.30 | Improved |
| add RF regularization on top of MMF (ablation) | TMMF · CIFAR10 (4000 labels) | top-1 error rate (%): 13.11 -> 10.90 | Improved |
| add MMF regularization on top of RF (ablation) | TRF · SVHN (250 labels) | top-1 error rate (%): 9.93 -> 5.02 | Improved |
| add MMF to RF with all labeled SVHN data (ablation) | TRF · SVHN (73257 labels) | top-1 error rate (%): 2.65 -> 2.42 | Improved |
| add MMF regularization to the plain transductive DCNN (ablation) | TDCNN · SVHN (500 labels) | top-1 error rate (%): 13.79 -> 7.23 | Improved |
| use unlabeled data via TSSDL instead of fully supervised training | Fully Supervised · CIFAR10 (1000 labels) | top-1 error rate (%): 46.43 -> 21.13 | Improved |
| combine TSSDL with Mean Teacher | TSSDL · CIFAR10 (1000 labels) | top-1 error rate (%): 21.13 -> 18.41 | Improved |
Closest papers by meaning
- TransMatch: A Transfer-Learning Scheme for Semi-Supervised Few-Shot Learning (CVPR 2020 · similarity 0.85)
- Label Propagation for Deep Semi-Supervised Learning (CVPR 2019 · similarity 0.84)
- A Realistic Evaluation of Semi-Supervised Learning for Fine-Grained Classification (CVPR 2021 · similarity 0.84)
- Realistic Evaluation of Deep Semi-Supervised Learning Algorithms (NeurIPS 2018 · similarity 0.84)
- TransBoost: Improving the Best ImageNet Performance using Deep Transduction (NeurIPS 2022 · similarity 0.84)
- Towards Realistic Semi-Supervised Learning (ECCV 2022 · similarity 0.84)
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.
