Research map / Label-efficient and robust learning / Semi-supervised learning
Generalized Semi-Supervised Learning via Self-Supervised Feature Adaptation
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 |
|---|---|---|---|
| Add self-supervised feature adaptation to FixMatch | FixMatch · CIFAR100, 400 labeled, ratio 0.5, UL | accuracy (%): 4.7 -> 23.2 | Helped |
| Add self-supervised feature adaptation to FixMatch | FixMatch · CIFAR100, 400 labeled, ratio 1.0, UL | accuracy (%): 3.5 -> 22.2 | Helped |
| Add self-supervised feature adaptation to FixMatch | FixMatch · CIFAR100, 4000 labeled, ratio 1.0, UL | accuracy (%): 16.0 -> 52.5 | Helped |
| Add self-supervised feature adaptation to FixMatch | FixMatch · CIFAR100, 10000 labeled, ratio 1.0, UL | accuracy (%): 33.0 -> 57.8 | Helped |
| Add self-supervised feature adaptation to FixMatch | FixMatch · OFFICE-31 A/D | accuracy (%): 46.9 -> 61.7 | Helped |
| Add self-supervised feature adaptation to FixMatch | FixMatch · OFFICE-HOME A/CPR | pseudo-label accuracy (%): 19.3 -> 69.0 | Helped |
| Replace rotation prediction with SimClr contrastive learning (ablation) | FM-SSFA · CIFAR100, 400 labeled, ratio 0.5, UL | accuracy (%): 23.2 -> 13.7 | Hurt |
| Replace rotation prediction with entropy minimization (ablation) | FM-SSFA · CIFAR100, 400 labeled, ratio 0.5, UL | accuracy (%): 23.2 -> 14.5 | Hurt |
| Remove the feature adaptation module, retaining rotation-task joint training (ablation) | FM-SSFA · OFFICE-31 and OFFICE-HOME | accuracy (%) | Hurt |
| Use 3 rather than 2 shared feature-extractor layers (ablation) | FM-SSFA · OFFICE-HOME | accuracy (%): 39.9 vs 39.8 | No clear effect |
| Share all 4 feature-extractor layers rather than 2 (ablation) | FM-SSFA · OFFICE-HOME | accuracy (%): 39.9 -> 39.1 | Hurt |
| Add SSFA in the low-mismatch setting | FixMatch · CIFAR100, ratio 0.0, L/UL | accuracy (%): 33.3 -> 41.3 | Helped |
Closest papers by meaning
- Boosting Semi-Supervised Learning by Exploiting All Unlabeled Data (CVPR 2023 · similarity 0.90)
- Bidirectional Adaptation for Robust Semi-Supervised Learning with Inconsistent Data Distributions (ICML 2023 · similarity 0.89)
- Bi-Level Optimization for Semi-Supervised Learning with Pseudo-Labeling (AAAI 2025 · similarity 0.89)
- Keep It on a Leash: Controllable Pseudo-label Generation Towards Realistic Long-Tailed Semi-Supervised Learning (NeurIPS 2025 · similarity 0.88)
- A Unified Framework for Heterogeneous Semi-supervised Learning (CVPR 2025 · similarity 0.87)
- SemiReward: A General Reward Model for Semi-supervised Learning (ICLR 2024 · similarity 0.87)
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.