适配方法

FlexMatch

介绍

在固定阈值的伪标签算法的基础上，FlexMatch提出课程伪标签法（CPL），实时估计不同类别的学习效果，动态调整阈值各类阈值，从而充分考虑不同类别样本之间的学习难度。

论文引用

@article{zhang2021flexmatch,
  title={FlexMatch: Boosting Semi-supervised Learning with Curriculum Pseudo Labeling},
  author={Zhang, Bowen and Wang, Yidong and Hou, Wenxin and Wu, Hao and Wang, Jindong and Okumura, Manabu and Shinozaki, Takahiro},
  booktitle={Neural Information Processing Systems (NeurIPS)},
  year={2021}
}

SimMatch

介绍

引入实例相似度，记录有标签数据的特征，使无标签数据与之对比，并且Teacher模型预测的实例相似度与分类结果相互矫正，从而获得更准确伪标签。

论文引用

@inproceedings{zheng2022simmatch,
  title={Simmatch: Semi-supervised learning with similarity matching},
  author={Zheng, Mingkai and You, Shan and Huang, Lang and Wang, Fei and Qian, Chen and Xu, Chang},
  booktitle={Computer Vision and Pattern Recognition},
  pages={14471--14481},
  year={2022}
}