Research map / Label-efficient and robust learning / Semi-supervised learning
Semi-supervised Learning with Multi-Head Co-Training
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 multi-head co-training with weak-to-strong augmentation and peer-generated pseudo-labels | WRN 28-2 · CIFAR-10, 4000 labels | error rate: 4.26±0.05 -> 3.84±0.09 | Helped |
| Use multi-head co-training with weak-to-strong augmentation and peer-generated pseudo-labels | WRN 28-2 · SVHN, 250 labels | error rate: 2.48±0.38 -> 2.21±0.18 | Helped |
| Use multi-head co-training with weak-to-strong augmentation and peer-generated pseudo-labels | WRN 28-8 · CIFAR-100, 10000 labels | error rate: 22.60±0.12 -> 21.68±0.16 | Helped |
| Use multi-head co-training with weak-to-strong augmentation and peer-generated pseudo-labels | ResNet-18 · Mini-ImageNet, 4000 labels | error rate: 46.32±0.27 -> 46.53±0.15 | No clear effect |
| Use multi-head co-training with weak-to-strong augmentation and peer-generated pseudo-labels | ResNet-18 · Mini-ImageNet, 10000 labels | error rate: 39.43±0.09 -> 39.74±0.12 | No clear effect |
| Add peer heads instead of using one head with self-generated pseudo-labels (ablation) | Multi-Head Co-Training · CIFAR-10, 4000 labels | error rate: 4.43 -> 3.84 | Helped |
| Give each head its own strong augmentation (ablation) | Multi-Head Co-Training · CIFAR-10, 4000 labels | error rate: 4.45 -> 3.84 | Helped |
| Add weak augmentation for pseudo-labeling instead of using original images (ablation) | Multi-Head Co-Training · CIFAR-10, 4000 labels | error rate: 4.86 -> 3.84 | Helped |
| Initialize heads independently rather than identically (ablation) | Multi-Head Co-Training · CIFAR-10, 4000 labels | error rate: 4.28 -> 3.84 | Helped |
| Add an exponential moving average model (ablation) | Multi-Head Co-Training · CIFAR-10, 4000 labels | error rate: 6.23 -> 3.84 | Helped |
Closest papers by meaning
- Learning by Association -- A Versatile Semi-Supervised Training Method for Neural Networks (CVPR 2017 · similarity 0.87)
- Three Heads Are Better than One: Complementary Experts for Long-Tailed Semi-supervised Learning (AAAI 2024 · similarity 0.85)
- CGMatch: A Different Perspective of Semi-supervised Learning (CVPR 2025 · similarity 0.85)
- Fast Co-Training under Weak Dependence via Stream-Based Active Learning (ICML 2024 · similarity 0.84)
- A Unified Framework for Heterogeneous Semi-supervised Learning (CVPR 2025 · similarity 0.83)
- Decouple then Classify: A Dynamic Multi-view Labeling Strategy with Shared and Specific Information (ICML 2024 · similarity 0.83)
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.