mlsanityData to 14 Oct 2026

MTSSL: Meta-Thresholding Semi-Supervised Learning

arXiv 2607.16363 ↗First posted 17 Jul 2026stat.ML cs.LGcited by 0

Not on the map. Why?

Abstract (arXiv)

A large body of Semi-supervised Learning~(SSL) algorithms encounter the threshold τ to select pseudo-labels. The value of τ across different SSL algorithms can vary depending on the learning perspective, yet they may achieve similar performance. It motivates us to establish a unified theoretical framework to explain the role of τ in SSL. We statistically explained that the unsupervised loss is affected independently by correct and incorrect pseudo-labels, while τ adjusts their numbers to balance the corresponding error term. This inherent trade-off indicates that SSL can reach the same loss with varying τ, precise optimal values of τ during training may be unnecessary. With this, we treat τ as an updatable parameter and optimize it via differentiation; the new policy is named Meta-Thresholding Semi-Supervised Learning (MTSSL). Extensive experiments demonstrate the superior performance of MTSSL. We observe that the accuracy curves of SSL algorithms can overlap completely even when the values of τ differ significantly, which supports our theoretical framework and indicates that the selection of τ can be relaxed in the future design of SSL algorithms.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 34 of its 50 references that are in the map. First posted on arXiv 17 Jul 2026.

Builds on

34 of 50 references.

References · 34

  1. 2014Explaining and Harnessing Adversarial Examplescited by 7,364
  2. 2021Convex optimizationcited by 2,515
  3. 2020FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidencecited by 2,085
  4. 2019MixMatch: A Holistic Approach to Semi-Supervised Learningcited by 1,391
  5. 2020Big Self-Supervised Models are Strong Semi-Supervised Learnerscited by 1,380
  6. 2019Unsupervised Data Augmentation for Consistency Trainingcited by 1,060
  7. 2017Virtual Adversarial Training: A Regularization Method for Supervised and Semi-Supervised Learningcited by 920
  8. 2020DivideMix: Learning with Noisy Labels as Semi-supervised Learningcited by 692
  9. 2021FlexMatch: Boosting Semi-Supervised Learning with Curriculum Pseudo Labelingcited by 540
  10. 2019Pseudo-Labeling and Confirmation Bias in Deep Semi-Supervised Learningcited by 510
  11. 2019ReMixMatch: Semi-Supervised Learning with Distribution Alignment and Augmentation Anchoringcited by 349
  12. 2019Label Propagation for Deep Semi-supervised Learningcited by 289
22 more
  1. 2015Semi-Supervised Learning with Ladder Networkscited by 247
  2. 2022FreeMatch: Self-adaptive Thresholding for Semi-supervised Learningcited by 232
  3. 2020Rethinking the Value of Labels for Improving Class-Imbalanced Learningcited by 216
  4. 2020CoMatch: Semi-supervised Learning with Contrastive Graph Regularizationcited by 167
  5. 2021Dash: Semi-Supervised Learning with Dynamic Thresholdingcited by 148
  6. 2023SoftMatch: Addressing the Quantity-Quality Trade-off in Semi-supervised Learningcited by 133
  7. 2022SimMatch: Semi-supervised Learning with Similarity Matchingcited by 118
  8. 2021AdaMatch: A Unified Approach to Semi-Supervised Learning and Domain Adaptationcited by 104
  9. 2021Semi-Supervised Domain Generalization with Stochastic StyleMatchcited by 43
  10. 2021Revisiting Consistency Regularization for Semi-Supervised Learningcited by 34
  11. 2017Low-shot learning with large-scale diffusioncited by 33
  12. 2023Boosting Semi-Supervised Learning by Exploiting All Unlabeled Datacited by 26
  13. 2023Shrinking Class Space for Enhanced Certainty in Semi-Supervised Learningcited by 17
  14. 2020Density-Aware Graph for Deep Semi-Supervised Visual Recognitioncited by 15
  15. 2021On Data-Augmentation and Consistency-Based Semi-Supervised Learningcited by 13
  16. 2023InstanT: Semi-supervised Learning with Instance-dependent Thresholdscited by 13
  17. 2023Boosting Semi-Supervised Learning by bridging high and low-confidence predictionscited by 11
  18. 2023MarginMatch: Improving Semi-Supervised Learning with Pseudo-Marginscited by 10
  19. 2020Semi-Supervised Learning with Meta-Gradientcited by 7
  20. 2020MLR-SNet: Transferable LR Schedules for Heterogeneous Taskscited by 5
  21. 2024RegMixMatch: Optimizing Mixup Utilization in Semi-Supervised Learningcited by 5
  22. 2023Exploration and Exploitation of Unlabeled Data for Open-Set Semi-Supervised Learningcited by 2

Built on by

Not cited yet.