Pattern Recognition, Год журнала: 2025, Номер unknown, С. 111365 - 111365
Опубликована: Янв. 1, 2025
Язык: Английский
Pattern Recognition, Год журнала: 2025, Номер unknown, С. 111365 - 111365
Опубликована: Янв. 1, 2025
Язык: Английский
Pattern Recognition, Год журнала: 2024, Номер 150, С. 110358 - 110358
Опубликована: Фев. 21, 2024
Язык: Английский
Процитировано
20Expert Systems with Applications, Год журнала: 2024, Номер 249, С. 123846 - 123846
Опубликована: Март 29, 2024
The collection of large-scale datasets inevitably introduces noisy labels, leading to a substantial degradation in the performance deep neural networks (DNNs). Although sample selection is mainstream method field learning with which aims mitigate impact labels during model training, testing these methods exhibits significant fluctuations across different noise rates and types. In this paper, we propose Cross-to-Merge Training (C2MT), novel framework that insensitive prior information progress, enhancing robustness. practical implementation, using cross-divided training data, two are cross-trained co-teaching strategy for several local rounds, subsequently merged into unified by performing federated averages on parameters models periodically. Additionally, introduce new class balance strategy, named Median Balance Strategy (MBS), cross-dividing process, evenly divides data labeled subset an unlabeled based estimated loss distribution characteristics. Extensive experimental results both synthetic real-world demonstrate effectiveness C2MT. Code will be available at: https://github.com/LanXiaoPang613/C2MT.
Язык: Английский
Процитировано
20Information Fusion, Год журнала: 2023, Номер 100, С. 101948 - 101948
Опубликована: Авг. 2, 2023
Язык: Английский
Процитировано
38Pattern Recognition, Год журнала: 2023, Номер 145, С. 109945 - 109945
Опубликована: Сен. 9, 2023
Язык: Английский
Процитировано
38Knowledge-Based Systems, Год журнала: 2023, Номер 278, С. 110817 - 110817
Опубликована: Июль 24, 2023
Язык: Английский
Процитировано
24IEEE Transactions on Neural Networks and Learning Systems, Год журнала: 2024, Номер 36(3), С. 5288 - 5303
Опубликована: Апрель 24, 2024
Multilabel feature selection solves the dimension distress of high-dimensional multilabel data by selecting optimal subset features. Noisy and incomplete labels raw hinder acquisition label-guided information. In existing approaches, mapping label space to a low-dimensional latent semantic decomposition mitigate noise is considered an effective strategy. However, decomposed contains redundant information, which misleads capture potential relevance. To eliminate effect information on extraction correlations, novel method named SLOFS via shared sublabel structure simultaneous orthogonal basis clustering for proposed. First, base (LOBSS) term engineered guide construction redundancy-free separated center structure. The LOBSS simultaneously retains Moreover, relevance nonredundant sublabels are fully explored. introduction graph regularization ensures structural consistency in sublabels, thus helping process. employs dynamic obtain high-quality uses constrain correlations projections. Finally, convergence provable optimization scheme proposed solve method. experimental studies 18 datasets demonstrate that presented performs consistently better than previous methods.
Язык: Английский
Процитировано
15Knowledge-Based Systems, Год журнала: 2024, Номер 297, С. 111948 - 111948
Опубликована: Май 18, 2024
Язык: Английский
Процитировано
12Knowledge-Based Systems, Год журнала: 2024, Номер 285, С. 111363 - 111363
Опубликована: Янв. 3, 2024
Язык: Английский
Процитировано
11Knowledge-Based Systems, Год журнала: 2024, Номер 290, С. 111566 - 111566
Опубликована: Фев. 24, 2024
Язык: Английский
Процитировано
10Information Sciences, Год журнала: 2023, Номер 648, С. 119525 - 119525
Опубликована: Авг. 16, 2023
Язык: Английский
Процитировано
23