MarginMatch: Improving Semi-Supervised Learning with Pseudo-Margins

被引:10
|
作者
Sosea, Tiberiu [1 ]
Caragea, Cornelia [1 ]
机构
[1] Univ Illinois, Chicago, IL 60680 USA
关键词
D O I
10.1109/CVPR52729.2023.01514
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
We introduce MarginMatch, a new SSL approach combining consistency regularization and pseudo-labeling, with its main novelty arising from the use of unlabeled data training dynamics to measure pseudo-label quality. Instead of using only the model's confidence on an unlabeled example at an arbitrary iteration to decide if the example should be masked or not, MarginMatch also analyzes the behavior of the model on the pseudo-labeled examples as the training progresses, to ensure low quality predictions are masked out. MarginMatch brings substantial improvements on four vision benchmarks in low data regimes and on two large-scale datasets, emphasizing the importance of enforcing high-quality pseudo-labels. Notably, we obtain an improvement in error rate over the state-of-the-art of 3.25% on CIFAR-100 with only 25 labels per class and of 3.78% on STL-10 using as few as 4 labels per class. We make our code available at https://github.com/tsosea2/MarginMatch.
引用
收藏
页码:15773 / 15782
页数:10
相关论文
共 50 条
  • [1] Improving Landmark Localization with Semi-Supervised Learning
    Honari, Sina
    Molchanov, Pavlo
    Tyree, Stephen
    Vincent, Pascal
    Pal, Christopher
    Kautz, Jan
    2018 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), 2018, : 1546 - 1555
  • [2] Pseudo Contrastive Learning for graph-based semi-supervised learning
    Lu, Weigang
    Guan, Ziyu
    Zhao, Wei
    Yang, Yaming
    Lv, Yuanhai
    Xing, Lining
    Yu, Baosheng
    Tao, Dacheng
    NEUROCOMPUTING, 2025, 624
  • [3] Improving semi-supervised learning through optimum connectivity
    Amorim, Willian P.
    Falcao, Alexandre X.
    Papa, Joao P.
    Carvalho, Marcelo H.
    PATTERN RECOGNITION, 2016, 60 : 72 - 85
  • [4] Pseudo-label semi-supervised learning for soybean monitoring
    Menezes, Gabriel Kirsten
    Astolfi, Gilberto
    Martins, Jose Augusto Correa
    Tetila, Everton Castelao
    Oliveira Jr, Adair da Silva
    Goncalves, Diogo Nunes
    Marcato Jr, Jose
    Silva, Jonathan Andrade
    Li, Jonathan
    Goncalves, Wesley Nunes
    Pistori, Hemerson
    SMART AGRICULTURAL TECHNOLOGY, 2023, 4
  • [5] Semi-supervised feature learning for improving writer identification
    Chen, Shiming
    Wang, Yisong
    Lin, Chin-Teng
    Ding, Weiping
    Cao, Zehong
    INFORMATION SCIENCES, 2019, 482 : 156 - 170
  • [6] Semi-Supervised Learning of Semantic Correspondence with Pseudo-Labels
    Kim, Jiwon
    Ryoo, Kwangrok
    Seo, Junyoung
    Lee, Gyuseong
    Kim, Daehwan
    Cho, Hansang
    Kim, Seungryong
    2022 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR 2022), 2022, : 19667 - 19677
  • [7] FlexMatch: Boosting Semi-Supervised Learning with Curriculum Pseudo Labeling
    Zhang, Bowen
    Wang, Yidong
    Hou, Wenxin
    Wu, Hao
    Wang, Jindong
    Okumura, Manabu
    Shinozaki, Takahiro
    ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 34 (NEURIPS 2021), 2021, 34
  • [8] Improving Semi-Supervised Learning for Audio Classification with FixMatch
    Grollmisch, Sascha
    Cano, Estefania
    ELECTRONICS, 2021, 10 (15)
  • [9] Semi-supervised Learning
    Adams, Niall
    JOURNAL OF THE ROYAL STATISTICAL SOCIETY SERIES A-STATISTICS IN SOCIETY, 2009, 172 : 530 - 530
  • [10] On semi-supervised learning
    A. Cholaquidis
    R. Fraiman
    M. Sued
    TEST, 2020, 29 : 914 - 937