Discriminator-free adversarial domain adaptation with information balance

In the realm of Unsupervised Domain Adaptation (UDA), adversarial learning has achieved significant progress. Existing adversarial UDA methods typically employ additional discriminators and feature extractors to engage in a max-min game. However, these methods often fail to effectively utilize the p...

Full description

Saved in:
Bibliographic Details
Main Authors: Hui Jiang, Di Wu, Xing Wei, Wenhao Jiang, Xiongbo Qing
Format: Article
Language:English
Published: AIMS Press 2025-01-01
Series:Electronic Research Archive
Subjects:
Online Access:https://www.aimspress.com/article/doi/10.3934/era.2025011
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850151447657709568
author Hui Jiang
Di Wu
Xing Wei
Wenhao Jiang
Xiongbo Qing
author_facet Hui Jiang
Di Wu
Xing Wei
Wenhao Jiang
Xiongbo Qing
author_sort Hui Jiang
collection DOAJ
description In the realm of Unsupervised Domain Adaptation (UDA), adversarial learning has achieved significant progress. Existing adversarial UDA methods typically employ additional discriminators and feature extractors to engage in a max-min game. However, these methods often fail to effectively utilize the predicted discriminative information, thus resulting in the mode collapse of the generator. In this paper, we propose a Dynamic Balance-based Domain Adaptation (DBDA) method for self-correlated domain adaptive image classification. Instead of adding extra discriminators, we repurpose the classifier as a discriminator and introduce a dynamic balancing learning approach. This approach ensures an explicit domain alignment and category distinction, thus enabling DBDA to fully leverage the predicted discriminative information for an effective feature alignment. We conducted experiments on multiple datasets, therefore demonstrating that the proposed method maintains a robust classification performance across various scenarios.
format Article
id doaj-art-b7ec14c6425f4dfead9c5c0eb82cef7c
institution OA Journals
issn 2688-1594
language English
publishDate 2025-01-01
publisher AIMS Press
record_format Article
series Electronic Research Archive
spelling doaj-art-b7ec14c6425f4dfead9c5c0eb82cef7c2025-08-20T02:26:15ZengAIMS PressElectronic Research Archive2688-15942025-01-0133121023010.3934/era.2025011Discriminator-free adversarial domain adaptation with information balanceHui Jiang0Di Wu1Xing Wei2Wenhao Jiang3Xiongbo Qing4Key Laboratory of Philosophy and Social Science of Anhui Province on Adolescent Mental Health and Crisis Intelligence Intervention, Hefei Normal University, Hefei 230601, ChinaKey Laboratory of Philosophy and Social Science of Anhui Province on Adolescent Mental Health and Crisis Intelligence Intervention, Hefei Normal University, Hefei 230601, ChinaSchool of Computer and Information, Hefei University of Technology, Hefei 230601, ChinaSchool of Computer and Information, Hefei University of Technology, Hefei 230601, ChinaSchool of Computer and Information, Hefei University of Technology, Hefei 230601, ChinaIn the realm of Unsupervised Domain Adaptation (UDA), adversarial learning has achieved significant progress. Existing adversarial UDA methods typically employ additional discriminators and feature extractors to engage in a max-min game. However, these methods often fail to effectively utilize the predicted discriminative information, thus resulting in the mode collapse of the generator. In this paper, we propose a Dynamic Balance-based Domain Adaptation (DBDA) method for self-correlated domain adaptive image classification. Instead of adding extra discriminators, we repurpose the classifier as a discriminator and introduce a dynamic balancing learning approach. This approach ensures an explicit domain alignment and category distinction, thus enabling DBDA to fully leverage the predicted discriminative information for an effective feature alignment. We conducted experiments on multiple datasets, therefore demonstrating that the proposed method maintains a robust classification performance across various scenarios.https://www.aimspress.com/article/doi/10.3934/era.2025011transfer learningdomain adaptationimage classificationdiscriminator-free adversarial learningdynamic balancing
spellingShingle Hui Jiang
Di Wu
Xing Wei
Wenhao Jiang
Xiongbo Qing
Discriminator-free adversarial domain adaptation with information balance
Electronic Research Archive
transfer learning
domain adaptation
image classification
discriminator-free adversarial learning
dynamic balancing
title Discriminator-free adversarial domain adaptation with information balance
title_full Discriminator-free adversarial domain adaptation with information balance
title_fullStr Discriminator-free adversarial domain adaptation with information balance
title_full_unstemmed Discriminator-free adversarial domain adaptation with information balance
title_short Discriminator-free adversarial domain adaptation with information balance
title_sort discriminator free adversarial domain adaptation with information balance
topic transfer learning
domain adaptation
image classification
discriminator-free adversarial learning
dynamic balancing
url https://www.aimspress.com/article/doi/10.3934/era.2025011
work_keys_str_mv AT huijiang discriminatorfreeadversarialdomainadaptationwithinformationbalance
AT diwu discriminatorfreeadversarialdomainadaptationwithinformationbalance
AT xingwei discriminatorfreeadversarialdomainadaptationwithinformationbalance
AT wenhaojiang discriminatorfreeadversarialdomainadaptationwithinformationbalance
AT xiongboqing discriminatorfreeadversarialdomainadaptationwithinformationbalance