Research map / Label-efficient and robust learning / Semi-supervised learning
RDA: Reciprocal Distribution Alignment for Robust Semi-Supervised Learning
Yue Duan, Lei Qi, Lei Wang, Luping Zhou, Yinghuan Shi
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 RDA instead of CoMatch with scarce labels | FixMatch-style SSL · CIFAR-10 (20 labels, matched) | accuracy: 88.43 -> 92.03 | Improved |
| use RDA instead of CoMatch | SSL · mini-ImageNet (1000 labels, matched) | accuracy: 43.72 -> 46.91 | Improved |
| use RDA instead of CoMatch with 100 labels | SSL · CIFAR-10 (100 labels, matched) | accuracy: 94.55 -> 94.35 | No clear effect |
| add original distribution alignment to FixMatch under imbalanced labeled data | FixMatch · CIFAR-10 (40 labels, imbalanced labeled N0=10) | accuracy: 85.72 -> 71.23 | Hurt |
| use RDA instead of FixMatch under imbalanced labeled data | FixMatch · CIFAR-10 (40 labels, imbalanced labeled N0=10) | accuracy: 85.72 -> 92.57 | Improved |
| use RDA instead of CoMatch under imbalanced labeled data | CoMatch · CIFAR-10 (40 labels, imbalanced labeled N0=10) | accuracy: 60.27 -> 92.57 | Improved |
| use RDA instead of FixMatch under imbalanced labeled data | FixMatch · CIFAR-100 (400 labels, imbalanced labeled N0=40) | accuracy: 25.66 -> 30.86 | Improved |
| use RDA instead of DARP | DARP protocol · CIFAR-10 DARP protocol (gamma_l=100, gamma_u=1) | accuracy: 85.40 -> 93.35 | Improved |
| use RDA instead of DARP | DARP protocol · STL-10 DARP protocol (gamma_l=10) | accuracy: 77.80 -> 87.21 | Improved |
| use reciprocal alignment of both directions instead of aligning only complementary predictions to reversed class predictions (ablation) | RDA · CIFAR-10 (40 labels, imbalanced labeled N0=20) | accuracy: 58.90 -> 81.78 | Improved |
| use reciprocal alignment of both directions instead of aligning only class predictions to reversed complementary predictions (ablation) | RDA · CIFAR-10 (40 labels, imbalanced labeled N0=20) | accuracy: 73.54 -> 81.78 | Improved |
| use reciprocal alignment in the conventional matched setting (ablation) | RDA · CIFAR-10 (40 labels, matched) | accuracy: 93.35 -> 94.13 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- FixMatch — FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidence (NeurIPS 2020) · better on 12 of 12 shared results
- CoMatch — CoMatch: Semi-Supervised Learning With Contrastive Graph Regularization (ICCV 2021) · better on 8 of 8 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
- DC-SSL: Addressing Mismatched Class Distribution in Semi-Supervised Learning (CVPR 2022 · similarity 0.90)
- Flexible Distribution Alignment: Towards Long-tailed Semi-supervised Learning with Proper Calibration (ECCV 2024 · similarity 0.90)
- Distribution Aligning Refinery of Pseudo-label for Imbalanced Semi-supervised Learning (NeurIPS 2020 · similarity 0.88)
- Bidirectional Adaptation for Robust Semi-Supervised Learning with Inconsistent Data Distributions (ICML 2023 · similarity 0.88)
- Three Heads Are Better than One: Complementary Experts for Long-Tailed Semi-supervised Learning (AAAI 2024 · similarity 0.86)
- A Square Peg in a Square Hole: Meta-Expert for Long-Tailed Semi-Supervised Learning (ICML 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.
