Research map / Label-efficient and robust learning / Semi-supervised learning
Unsupervised Semantic Aggregation and Deformable Template Matching for Semi-Supervised Learning
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 T-MI instead of single-pair IIC mutual information | Unsupervised clustering model · MNIST | accuracy: 95.93% -> 98.82% | Helped |
| Replace FixMatch confidence label generator with DTM | FixMatch framework · CIFAR-10, 40/250/4000 labels | accuracy: 88.51 -> 89.48; 94.16 -> 94.49; 94.5 -> 95.2 | Helped |
| Add DTM to supervised learning with few labeled samples (ablation) | CIFAR-10 labeled baseline · CIFAR-10, 250 labels | accuracy: 45.52 -> 94.27 | Helped |
| Add DTM to supervised learning with few labeled samples (ablation) | CIFAR-10 labeled baseline · CIFAR-10, 4000 labels | accuracy: 83.97 -> 95.01 | Helped |
| Add T-MI semantic aggregation (ablation) | CIFAR-10 labeled baseline · CIFAR-10, 250 labels | accuracy: 45.52 -> 81.23 | Helped |
| Add T-MI semantic aggregation (ablation) | CIFAR-10 labeled baseline · CIFAR-10, 4000 labels | accuracy: 83.97 -> 88.34 | Helped |
| Add the C×K additional feature samples to build class centers (ablation) | USADTM · CIFAR-10, 250 and 4000 labels | accuracy: 94.58 -> 95.21; 95.22 -> 95.74 | Helped |
| Set the cosine similarity threshold to τ=0.85 rather than 0.95 or 0.90 (ablation) | USADTM · CIFAR-10, 250 and 4000 labels | accuracy: 93.45 -> 95.21; 94.49 -> 95.74; 94.02 -> 95.21; 95.16 -> 95.74 | Helped |
| Set the cosine similarity threshold to τ=0.85 rather than 0.80 (ablation) | USADTM · CIFAR-10, 250 and 4000 labels | accuracy: 95.02 -> 95.21; 95.45 -> 95.74 | Mixed |
| Increase T-MI loss weight from α=0.1 to α=1 (ablation) | USADTM · CIFAR-10, 250 and 4000 labels | accuracy: 95.21 -> 93.68; 95.74 -> 94.07 | Hurt |
| Reduce T-MI loss weight from α=0.1 to α=0.01 (ablation) | USADTM · CIFAR-10, 250 and 4000 labels | accuracy: 95.21 -> 94.86; 95.74 -> 95.14 | Hurt |
Closest papers by meaning
- Scaling Up Semi-supervised Learning with Unconstrained Unlabelled Data (AAAI 2024 · similarity 0.88)
- InterLUDE: Interactions between Labeled and Unlabeled Data to Enhance Semi-Supervised Learning (ICML 2024 · similarity 0.87)
- A Unified Framework for Heterogeneous Semi-supervised Learning (CVPR 2025 · similarity 0.87)
- Learning with Partial-Label and Unlabeled Data: A Uniform Treatment for Supervision Redundancy and Insufficiency (ICML 2024 · similarity 0.86)
- OwMatch: Conditional Self-Labeling with Consistency for Open-World Semi-Supervised Learning (NeurIPS 2024 · similarity 0.86)
- CGMatch: A Different Perspective of Semi-supervised Learning (CVPR 2025 · similarity 0.85)
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.