Uncertainty guided ensemble self-training for semi-supervised global field reconstruction

Autor: Yunyang Zhang, Zhiqiang Gong, Xiaoyu Zhao, Wen Yao
Jazyk: angličtina
Rok vydání: 2023
Předmět:
Zdroj: Complex & Intelligent Systems, Vol 10, Iss 1, Pp 469-483 (2023)
Druh dokumentu: article
ISSN: 2199-4536
2198-6053
DOI: 10.1007/s40747-023-01167-4
Popis: Abstract Recovering the global accurate complex physics field from limited sensors is critical to the measurement and control of the engineering system. General reconstruction methods for recovering the field, especially the deep learning with more parameters and better representational ability, usually require large amounts of labeled data which is unaffordable in practice. To solve the problem, this paper proposes uncertainty guided ensemble self-training (UGE-ST), using plentiful unlabeled data to improve reconstruction performance and reduce the required labeled data. A novel self-training framework with the ensemble teacher and pre-training student designed to improve the accuracy of the pseudo-label and remedy the impact of noise is first proposed. On the other hand, uncertainty guided learning is proposed to encourage the model to focus on the highly confident regions of pseudo-labels and mitigate the effects of wrong pseudo-labeling in self-training, improving the performance of the reconstruction model. Experiments including the airfoil velocity and pressure field reconstruction and the electronic components’ temperature field reconstruction indicate that our UGE-ST can save up to 90% of the data with the same accuracy as supervised learning.
Databáze: Directory of Open Access Journals
Externí odkaz: