Research map / Label-efficient and robust learning / Semi-supervised learning
Consistency-based Semi-supervised Active Learning: Towards Minimizing Labeling Cost
Mingfei Gao, Zizhao Zhang, Guo Yu, Sercan . Arık, Larry S. Davis, Tomas Pfister
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 |
|---|---|---|---|
| use consistency-based selection instead of uniform sampling under SSL | MixMatch AL · CIFAR-10 (250 labels, SSL) | accuracy: 87.78 -> 90.23 | Improved |
| use consistency-based selection instead of k-center under SSL | MixMatch AL · CIFAR-10 (250 labels, SSL) | accuracy: 88.75 -> 90.23 | Improved |
| use consistency-based selection instead of uniform at 4000 labels | MixMatch AL · CIFAR-10 (4000 labels, SSL) | accuracy: 93.45 -> 94.57 | Improved |
| entropy selection instead of uniform under SSL | MixMatch AL · CIFAR-10 (500 labels, SSL) | accuracy: 90.50 -> 89.95 | Hurt |
| add SSL training to AL (vs supervised baselines) | active learning · CIFAR-10 (150 labels) | accuracy: 48.33 -> 87.57 | Improved |
| vary AL batch size 25, 50, 100 (ablation) | consistency-based SSL AL · CIFAR-10 (200 labels) | accuracy: 89.5, 89.2, 89.3 | No clear effect |
| consistency selection in supervised setting vs uniform | consistency sampling · CIFAR-10 supervised (3000 labels) | accuracy: 80.81 -> 82.75 | No clear effect |
| start with K0=50 instead of 100 labeled samples (ablation) | consistency-based SSL AL · CIFAR-10 | accuracy at 150 labels | Hurt |
| start with K0=100 instead of 150 or 200 (ablation) | consistency-based SSL AL · CIFAR-10 | accuracy under fixed budget | Improved |
| add SSL training and consistency selection vs best AL baseline VAAL at 40% labels | consistency-based SSL AL · CIFAR-100 | accuracy: +22.62% accuracy | Improved |
| same comparison on ImageNet at 40% labels | consistency-based SSL AL · ImageNet | accuracy: +12.28% accuracy | Improved |
Closest papers by meaning
- Towards Cost-Effective Learning: A Synergy of Semi-Supervised and Active Learning (CVPR 2025 · similarity 0.93)
- CGMatch: A Different Perspective of Semi-supervised Learning (CVPR 2025 · similarity 0.89)
- Composite Active Learning: Towards Multi-Domain Active Learning with Theoretical Guarantees (AAAI 2024 · similarity 0.89)
- Enhancing Semi-Supervised Learning via Representative and Diverse Sample Selection (NeurIPS 2024 · similarity 0.88)
- InterLUDE: Interactions between Labeled and Unlabeled Data to Enhance Semi-Supervised Learning (ICML 2024 · similarity 0.88)
- Instance-wise Supervision-level Optimization in Active Learning (CVPR 2025 · similarity 0.88)
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.
