Research map / Label-efficient and robust learning / Semi-supervised learning
Exploration and Exploitation of Unlabeled Data for Open-Set Semi-supervised Learning
Ganlong Zhao, Guanbin Li, Yipeng Qin, Jinjin Zhang, Zhenhua Chai, Xiaolin K. Wei, Liang Lin, Yizhou Yu
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 prototype clustering and identification (ablation) | FixMatch · DomainNet-Real ID10/OOD50k | accuracy: 52.8 -> 55.0 | Improved |
| add prototype clustering and identification (ablation) | FixMatch · DomainNet-Real ID20/OOD50k | accuracy: 49.7 -> 52.5 | Improved |
| cluster strong augmentations as well as weak ones (ablation) | clustering · DomainNet-Real ID10/OOD50k | accuracy: 54.2 -> 55.0 | Improved |
| add cascading-pool importance sampling on top of clustering (ablation) | Ours · DomainNet-Real ID10/OOD50k | accuracy: 55.0 -> 59.4 | Improved |
| add cascading-pool importance sampling on top of clustering (ablation) | Ours · DomainNet-Real ID20/OOD50k | accuracy: 52.5 -> 54.3 | Improved |
| use importance-based selection instead of random selection of ID samples (ablation) | Ours · DomainNet-Real ID10/OOD50k | accuracy: 57.2 -> 58.1 | Improved |
| add cascading pools to importance sampling (ablation) | Ours · DomainNet-Real ID20/OOD50k | accuracy: 53.7 -> 54.3 | Improved |
| use two cascaded pools instead of zero pools (ablation) | Ours · DomainNet-Real ID20/OOD50k | accuracy: 52.5 -> 54.3 | Improved |
| exploit OOD samples with SimCLR loss instead of masking them (ablation) | FixMatch · DomainNet-Real ID20/OOD50k | accuracy: 54.3 -> 56.7 | Improved |
| mask OOD samples in FixMatch loss | FixMatch · DomainNet-Real ID20/OOD50k | accuracy: 49.7 -> 54.3 | Improved |
| train with unlabeled OOD samples only (clustering) instead of labeled only | Ours (Clustering) · CIFAR-100 ID10/OOD90 (extreme case) | accuracy: 47.3 -> 73.5 | Improved |
| integrate the method into UDA | UDA · CIFAR-100 ID10/OOD90 | accuracy: 38.9 -> 48.4 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- FixMatch — FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidence (NeurIPS 2020) · better on 7 of 7 shared results
- OpenMatch — OpenMatch: Open-Set Semi-supervised Learning with Open-set Consistency Regularization (NeurIPS 2021) · better on 6 of 6 shared results
- FlexMatch — FlexMatch: Boosting Semi-Supervised Learning with Curriculum Pseudo Labeling (NeurIPS 2021) · better on 1 of 1 shared result
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
- Multi-Task Curriculum Framework for Open-Set Semi-Supervised Learning (ECCV 2020 · similarity 0.92)
- Rethinking Safe Semi-supervised Learning: Transferring the Open-set Problem to A Close-set One (ICCV 2023 · similarity 0.91)
- STEP: Out-of-Distribution Detection in the Presence of Limited In-Distribution Labeled Data (NeurIPS 2021 · similarity 0.91)
- Trash To Treasure: Harvesting OOD Data With Cross-Modal Matching for Open-Set Semi-Supervised Learning (ICCV 2021 · similarity 0.91)
- Binary Decomposition: A Problem Transformation Perspective for Open-Set Semi-Supervised Learning (ICML 2024 · similarity 0.87)
- How Does Unlabeled Data Provably Help Out-of-Distribution Detection? (ICLR 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.
