Research map / Label-efficient and robust learning / Semi-supervised learning
SSB: Simple but Strong Baseline for Boosting Performance of Open-Set Semi-Supervised Learning
Yue Fan, Anna Kukleva, Dengxin Dai, Bernt Schiele
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-based pseudo-labeling on all unlabeled data (inliers and OOD) for classifier training (ablation) | SSB · CIFAR-10 (25 labels) | inlier accuracy (%): 78.47 -> 91.65 | Improved |
| add confidence filtering without projection heads (ablation) | SSB · CIFAR-10 (25 labels) | outlier AUROC (%): 89.67 -> 63.46 | Hurt |
| add two separate MLP projection heads (with confidence filtering) (ablation) | SSB · CIFAR-10 (25 labels) | outlier AUROC (%): 63.46 -> 94.76 | Improved |
| use a shared projection head instead of two separate heads (ablation) | SSB · CIFAR-10 (25 labels) | outlier AUROC (%): 94.76 -> 63.87 | Hurt |
| use confidence-based pseudo-labeling instead of detector-based filtering (ablation) | SSB · CIFAR-10 (25 labels) | inlier accuracy (%): 47.20 -> 91.65 | Improved |
| keep pseudo-labeled OOD data instead of removing it with an oracle (ablation) | SSB · CIFAR-10 (25 labels) | inlier accuracy (%): 88.28 -> 91.65 | Improved |
| add pseudo-negative mining for detector training (ablation) | SSB · CIFAR-10 (25 labels) | outlier AUROC (%): 90.03 -> 94.76 | Improved |
| use standard pseudo-labeling for the detector instead of pseudo-negative mining (ablation) | SSB · CIFAR-10 (25 labels) | outlier AUROC (%): 90.03 -> 89.69 | No clear effect |
| vary pseudo-negative threshold theta from 0.2 to 0.001 (ablation) | SSB · CIFAR-10 (25 labels) | outlier seen AUROC (%): 92.96 - 94.76 | No clear effect |
| use OVA classifiers as outlier detector instead of binary classifier or cross-modal matching (ablation) | SSB · CIFAR-100 (55 inlier classes) | outlier AUROC (%): 76.12 -> 82.62 | Improved |
| add SSB components to existing open-set SSL methods | MTC / T2T / OpenMatch · CIFAR-100 (55 inlier classes, 25 labels/class) | outlier AUROC (%): 52.93 -> 69.50 (T2T) | Improved |
| add MLP heads to other methods for equal parameter count (ablation) | baselines · CIFAR datasets | accuracy / AUROC | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- OpenMatch — OpenMatch: Open-Set Semi-supervised Learning with Open-set Consistency Regularization (NeurIPS 2021) · better on 4 of 5 shared results
- FixMatch — FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidence (NeurIPS 2020) · better on 1 of 2 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
- IOMatch: Simplifying Open-Set Semi-Supervised Learning with Joint Inliers and Outliers Utilization (ICCV 2023 · similarity 0.94)
- Binary Decomposition: A Problem Transformation Perspective for Open-Set Semi-Supervised Learning (ICML 2024 · similarity 0.90)
- Unknown-Aware Graph Regularization for Robust Semi-supervised Learning from Uncurated Data (AAAI 2024 · similarity 0.88)
- Re-Evaluating the Impact of Unseen-Class Unlabeled Data on Semi-Supervised Learning Model (ICLR 2025 · similarity 0.87)
- ANEDL: Adaptive Negative Evidential Deep Learning for Open-Set Semi-supervised Learning (AAAI 2024 · similarity 0.86)
- Rethinking Confidence Scores and Thresholds in Pseudolabeling-based SSL (ICML 2025 · 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.
