MeTa Learning-Based Optimization of Unsupervised Domain Adaptation Deep Networks
This paper introduces a novel unsupervised domain adaptation (UDA) method, MeTa Discriminative Class-Wise MMD (MCWMMD), which combines meta-learning with a Class-Wise Maximum Mean Discrepancy (MMD) approach to enhance domain adaptation. Traditional MMD methods align overall distributions but struggl...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2025-01-01
|
| Series: | Mathematics |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2227-7390/13/2/226 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832588068519936000 |
|---|---|
| author | Hsiau-Wen Lin Trang-Thi Ho Ching-Ting Tu Hwei-Jen Lin Chen-Hsiang Yu |
| author_facet | Hsiau-Wen Lin Trang-Thi Ho Ching-Ting Tu Hwei-Jen Lin Chen-Hsiang Yu |
| author_sort | Hsiau-Wen Lin |
| collection | DOAJ |
| description | This paper introduces a novel unsupervised domain adaptation (UDA) method, MeTa Discriminative Class-Wise MMD (MCWMMD), which combines meta-learning with a Class-Wise Maximum Mean Discrepancy (MMD) approach to enhance domain adaptation. Traditional MMD methods align overall distributions but struggle with class-wise alignment, reducing feature distinguishability. MCWMMD incorporates a meta-module to dynamically learn a deep kernel for MMD, improving alignment accuracy and model adaptability. This meta-learning technique enhances the model’s ability to generalize across tasks by ensuring domain-invariant and class-discriminative feature representations. Despite the complexity of the method, including the need for meta-module training, it presents a significant advancement in UDA. Future work will explore scalability in diverse real-world scenarios and further optimize the meta-learning framework. MCWMMD offers a promising solution to the persistent challenge of domain adaptation, paving the way for more adaptable and generalizable deep learning models. |
| format | Article |
| id | doaj-art-7e7efb0201ae4a33a7703a0054127293 |
| institution | Kabale University |
| issn | 2227-7390 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Mathematics |
| spelling | doaj-art-7e7efb0201ae4a33a7703a00541272932025-01-24T13:39:48ZengMDPI AGMathematics2227-73902025-01-0113222610.3390/math13020226MeTa Learning-Based Optimization of Unsupervised Domain Adaptation Deep NetworksHsiau-Wen Lin0Trang-Thi Ho1Ching-Ting Tu2Hwei-Jen Lin3Chen-Hsiang Yu4Department of Information Management, Chihlee University of Technology, Taipei 220305, TaiwanDepartment of Computer Science and Information Engineering, Tamkang University, Taipei 251301, TaiwanDepartment of Applied Mathematics, National Chung Hsing University, Taichung 402202, TaiwanDepartment of Computer Science and Information Engineering, Tamkang University, Taipei 251301, TaiwanMultidisciplinary Graduate Engineering, College of Engineering, Northeastern University, Boston, MA 02115, USAThis paper introduces a novel unsupervised domain adaptation (UDA) method, MeTa Discriminative Class-Wise MMD (MCWMMD), which combines meta-learning with a Class-Wise Maximum Mean Discrepancy (MMD) approach to enhance domain adaptation. Traditional MMD methods align overall distributions but struggle with class-wise alignment, reducing feature distinguishability. MCWMMD incorporates a meta-module to dynamically learn a deep kernel for MMD, improving alignment accuracy and model adaptability. This meta-learning technique enhances the model’s ability to generalize across tasks by ensuring domain-invariant and class-discriminative feature representations. Despite the complexity of the method, including the need for meta-module training, it presents a significant advancement in UDA. Future work will explore scalability in diverse real-world scenarios and further optimize the meta-learning framework. MCWMMD offers a promising solution to the persistent challenge of domain adaptation, paving the way for more adaptable and generalizable deep learning models.https://www.mdpi.com/2227-7390/13/2/226unsupervised domain adaptationmaximum mean discrepancy (MMD)discriminative class-wise MMD (DCWMMD)meta-learningdeep kernelfeature distributions |
| spellingShingle | Hsiau-Wen Lin Trang-Thi Ho Ching-Ting Tu Hwei-Jen Lin Chen-Hsiang Yu MeTa Learning-Based Optimization of Unsupervised Domain Adaptation Deep Networks Mathematics unsupervised domain adaptation maximum mean discrepancy (MMD) discriminative class-wise MMD (DCWMMD) meta-learning deep kernel feature distributions |
| title | MeTa Learning-Based Optimization of Unsupervised Domain Adaptation Deep Networks |
| title_full | MeTa Learning-Based Optimization of Unsupervised Domain Adaptation Deep Networks |
| title_fullStr | MeTa Learning-Based Optimization of Unsupervised Domain Adaptation Deep Networks |
| title_full_unstemmed | MeTa Learning-Based Optimization of Unsupervised Domain Adaptation Deep Networks |
| title_short | MeTa Learning-Based Optimization of Unsupervised Domain Adaptation Deep Networks |
| title_sort | meta learning based optimization of unsupervised domain adaptation deep networks |
| topic | unsupervised domain adaptation maximum mean discrepancy (MMD) discriminative class-wise MMD (DCWMMD) meta-learning deep kernel feature distributions |
| url | https://www.mdpi.com/2227-7390/13/2/226 |
| work_keys_str_mv | AT hsiauwenlin metalearningbasedoptimizationofunsuperviseddomainadaptationdeepnetworks AT trangthiho metalearningbasedoptimizationofunsuperviseddomainadaptationdeepnetworks AT chingtingtu metalearningbasedoptimizationofunsuperviseddomainadaptationdeepnetworks AT hweijenlin metalearningbasedoptimizationofunsuperviseddomainadaptationdeepnetworks AT chenhsiangyu metalearningbasedoptimizationofunsuperviseddomainadaptationdeepnetworks |