Research map / Label-efficient and robust learning / Semi-supervised learning
Credal Self-Supervised Learning
Julian Lienen, Eyke Hüllermeier
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 |
|---|---|---|---|
| replace hard thresholded pseudo-labels with credal target sets | FixMatch · CIFAR-10 40 labels | misclassification rate (%): 11.39 -> 6.50 | Improved |
| replace hard thresholded pseudo-labels with credal target sets | FixMatch · SVHN 40 labels | misclassification rate (%): 7.65 -> 3.67 | Improved |
| replace hard thresholded pseudo-labels with credal target sets | FixMatch · CIFAR-100 400 labels | misclassification rate (%): 49.95 -> 43.43 | Improved |
| replace hard thresholded pseudo-labels with credal target sets | FixMatch · CIFAR-10 4000 labels | misclassification rate (%): 4.31 -> 4.43 | No clear effect |
| replace hard thresholded pseudo-labels with credal target sets | FixMatch · CIFAR-10 250 labels | misclassification rate (%): 5.07 -> 5.48 | Hurt |
| replace hard thresholded pseudo-labels with credal target sets | FixMatch · SVHN 1000 labels | misclassification rate (%): 2.28 -> 1.99 | Improved |
| replace hard thresholded pseudo-labels with credal target sets | FixMatch · CIFAR-10 40 labels | ECE: 0.087 -> 0.032 | Improved |
| replace hard thresholded pseudo-labels with credal target sets | FixMatch · SVHN 1000 labels | ECE: 0.010 -> 0.007 | Improved |
| use credal sets instead of adaptive label smoothing | LSMatch · SVHN 40 labels | ECE: 0.086 -> 0.022 | Improved |
| train for 1/8 of the iterations with credal targets (no threshold) | CSSL · CIFAR-10 40 labels | misclassification rate (%): 14.73 -> 10.04 | Improved |
| lower FixMatch confidence threshold from 0.95 to 0.8 | FixMatch · CIFAR-10 40 labels, 1/8 iterations | misclassification rate (%): 14.73 -> 11.99 | Improved |
| set FixMatch confidence threshold to 0.0 | FixMatch · CIFAR-10 40 labels, 1/8 iterations | misclassification rate (%): 11.99 -> 18.50 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 3 on most shared results.
- FixMatch — FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidence (NeurIPS 2020) · better on 5 of 5 shared results
- AlphaMatch — AlphaMatch: Improving Consistency for Semi-Supervised Learning With Alpha-Divergence (CVPR 2021) · better on 1 of 3 shared results
- CoMatch — CoMatch: Semi-Supervised Learning With Contrastive Graph Regularization (ICCV 2021) · better on 1 of 2 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
- Credal Prediction based on Relative Likelihood (NeurIPS 2025 · similarity 0.86)
- Credal Learning Theory (NeurIPS 2024 · similarity 0.86)
- Conformalized Credal Set Predictors (NeurIPS 2024 · similarity 0.86)
- OwMatch: Conditional Self-Labeling with Consistency for Open-World Semi-Supervised Learning (NeurIPS 2024 · similarity 0.84)
- Self-supervised Representation Learning from Random Data Projectors (ICLR 2024 · similarity 0.84)
- Keep It on a Leash: Controllable Pseudo-label Generation Towards Realistic Long-Tailed Semi-Supervised Learning (NeurIPS 2025 · similarity 0.84)
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.
