Research map / Label-efficient and robust learning / Semi-supervised learning
FeatMatch: Feature-Based Augmentation for Semi-Supervised Learning
Chia-Wen Kuo, Chih-Yao Ma, Jia-Bin Huang, Zsolt Kira
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 feature-based augmentation (AugF) with both consistency losses (ablation) | image-based baseline · CIFAR-10, 250 labels | error rate (%): 19.55 -> 7.90 | Improved |
| add feature-based augmentation (AugF) with both consistency losses (ablation) | image-based baseline · CIFAR-10, 4,000 labels | error rate (%): 6.08 -> 5.00 | Improved |
| add L_con-f (ablation) | FeatMatch w/o L_con-f · CIFAR-10, 250 labels | error rate (%): 18.57 -> 7.90 | Improved |
| add L_con-f (ablation) | FeatMatch w/o L_con-f · CIFAR-10, 4,000 labels | error rate (%): 6.09 -> 5.00 | Improved |
| add L_con-g (ablation) | FeatMatch w/o L_con-g · CIFAR-10, 250 labels | error rate (%): 8.19 -> 7.90 | Improved |
| add L_con-g (ablation) | FeatMatch w/o L_con-g · CIFAR-10, 4,000 labels | error rate (%): 5.16 -> 5.00 | Improved |
| use FeatMatch instead of PLCB | PLCB · mini-ImageNet, 4,000 labels | error rate (%): 56.49 -> 39.05 | Improved |
| use FeatMatch instead of PLCB | PLCB · CIFAR-100, 4,000 labels | error rate (%): 37.55 -> 31.06 | Improved |
| add feature-based augmentation under domain shift | semi-supervised baseline · DomainNet, r_u=50% | error rate (%): 65.82 -> 54.01 | Improved |
| add feature-based augmentation with clean unlabeled data | semi-supervised baseline · DomainNet, r_u=0% | error rate (%): 56.63 -> 40.66 | Improved |
| increase share of shifted-domain unlabeled data from 0% to 50% | FeatMatch · DomainNet | error rate (%): 40.66 -> 54.01 | Hurt |
| add distribution alignment and Cutout | FeatMatch · CIFAR-10, 250 labels | error rate (%): 7.50 -> 6.00 | Improved |
Later papers that compared with this method
1 later accepted paper reports FeatMatch as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Revisiting Consistency Regularization for Semi-Supervised Learning (IJCV 2022 · better on 2 of 2 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- ReMixMatch — ReMixMatch: Semi-Supervised Learning with Distribution Matching and Augmentation Anchoring (ICLR 2020) · better on 1 of 6 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
- ConMatch: Semi-Supervised Learning with Confidence-Guided Consistency Regularization (ECCV 2022 · similarity 0.90)
- AlphaMatch: Improving Consistency for Semi-Supervised Learning With Alpha-Divergence (CVPR 2021 · similarity 0.89)
- Data Augmentation with Diffusion for Open-Set Semi-Supervised Learning (NeurIPS 2024 · similarity 0.87)
- Scaling Up Semi-supervised Learning with Unconstrained Unlabelled Data (AAAI 2024 · similarity 0.86)
- OTMatch: Improving Semi-Supervised Learning with Optimal Transport (ICML 2024 · similarity 0.86)
- Towards Generalizing to Unseen Domains with Few Labels (CVPR 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.
