Unsupervised Domain Adaptation Method Based on Relative Entropy Regularization and Measure Propagation

This paper presents a novel unsupervised domain adaptation (UDA) framework that integrates information-theoretic principles to mitigate distributional discrepancies between source and target domains. The proposed method incorporates two key components: (1) relative entropy regularization, which leve...

Full description

Saved in:
Bibliographic Details
Main Authors: Lianghao Tan, Zhuo Peng, Yongjia Song, Xiaoyi Liu, Huangqi Jiang, Shubing Liu, Weixi Wu, Zhiyuan Xiang
Format: Article
Language:English
Published: MDPI AG 2025-04-01
Series:Entropy
Subjects:
Online Access:https://www.mdpi.com/1099-4300/27/4/426
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849713701499699200
author Lianghao Tan
Zhuo Peng
Yongjia Song
Xiaoyi Liu
Huangqi Jiang
Shubing Liu
Weixi Wu
Zhiyuan Xiang
author_facet Lianghao Tan
Zhuo Peng
Yongjia Song
Xiaoyi Liu
Huangqi Jiang
Shubing Liu
Weixi Wu
Zhiyuan Xiang
author_sort Lianghao Tan
collection DOAJ
description This paper presents a novel unsupervised domain adaptation (UDA) framework that integrates information-theoretic principles to mitigate distributional discrepancies between source and target domains. The proposed method incorporates two key components: (1) relative entropy regularization, which leverages Kullback–Leibler (KL) divergence to align the predicted label distribution of the target domain with a reference distribution derived from the source domain, thereby reducing prediction uncertainty; and (2) measure propagation, a technique that transfers probability mass from the source domain to generate pseudo-measures—estimated probabilistic representations—for the unlabeled target domain. This dual mechanism enhances both global feature alignment and semantic consistency across domains. Extensive experiments on benchmark datasets (OfficeHome and DomainNet) demonstrate that the proposed approach consistently outperforms State-of-the-Art methods, particularly in scenarios with significant domain shifts. These results confirm the robustness, scalability, and theoretical grounding of our framework, offering a new perspective on the fusion of information theory and domain adaptation.
format Article
id doaj-art-845dddcdb493489186fdb591adc82f2f
institution DOAJ
issn 1099-4300
language English
publishDate 2025-04-01
publisher MDPI AG
record_format Article
series Entropy
spelling doaj-art-845dddcdb493489186fdb591adc82f2f2025-08-20T03:13:54ZengMDPI AGEntropy1099-43002025-04-0127442610.3390/e27040426Unsupervised Domain Adaptation Method Based on Relative Entropy Regularization and Measure PropagationLianghao Tan0Zhuo Peng1Yongjia Song2Xiaoyi Liu3Huangqi Jiang4Shubing Liu5Weixi Wu6Zhiyuan Xiang7Department of Computer Science, Arizona State University, Tempe, AZ 85281, USADepartment of Computer Science, Arizona State University, Tempe, AZ 85281, USADepartment of Language Science, University of California, Irvine, CA 92697, USADepartment of Computer Science, Arizona State University, Tempe, AZ 85281, USADepartment of Computer Science, Georgia Institute of Technology, Atlanta, GA 30332, USADepartment of Computer Science, North Carolina at Chapel Hill, Orange, GA 27599, USADepartment of Computer Science, New York University, Brooklyn, NY 10003, USADepartment of Computer Science, University of California, San Diego, CA 92093, USAThis paper presents a novel unsupervised domain adaptation (UDA) framework that integrates information-theoretic principles to mitigate distributional discrepancies between source and target domains. The proposed method incorporates two key components: (1) relative entropy regularization, which leverages Kullback–Leibler (KL) divergence to align the predicted label distribution of the target domain with a reference distribution derived from the source domain, thereby reducing prediction uncertainty; and (2) measure propagation, a technique that transfers probability mass from the source domain to generate pseudo-measures—estimated probabilistic representations—for the unlabeled target domain. This dual mechanism enhances both global feature alignment and semantic consistency across domains. Extensive experiments on benchmark datasets (OfficeHome and DomainNet) demonstrate that the proposed approach consistently outperforms State-of-the-Art methods, particularly in scenarios with significant domain shifts. These results confirm the robustness, scalability, and theoretical grounding of our framework, offering a new perspective on the fusion of information theory and domain adaptation.https://www.mdpi.com/1099-4300/27/4/426unsupervised domain adaptationinformation theoryrelative entropy regularizationprobability measure
spellingShingle Lianghao Tan
Zhuo Peng
Yongjia Song
Xiaoyi Liu
Huangqi Jiang
Shubing Liu
Weixi Wu
Zhiyuan Xiang
Unsupervised Domain Adaptation Method Based on Relative Entropy Regularization and Measure Propagation
Entropy
unsupervised domain adaptation
information theory
relative entropy regularization
probability measure
title Unsupervised Domain Adaptation Method Based on Relative Entropy Regularization and Measure Propagation
title_full Unsupervised Domain Adaptation Method Based on Relative Entropy Regularization and Measure Propagation
title_fullStr Unsupervised Domain Adaptation Method Based on Relative Entropy Regularization and Measure Propagation
title_full_unstemmed Unsupervised Domain Adaptation Method Based on Relative Entropy Regularization and Measure Propagation
title_short Unsupervised Domain Adaptation Method Based on Relative Entropy Regularization and Measure Propagation
title_sort unsupervised domain adaptation method based on relative entropy regularization and measure propagation
topic unsupervised domain adaptation
information theory
relative entropy regularization
probability measure
url https://www.mdpi.com/1099-4300/27/4/426
work_keys_str_mv AT lianghaotan unsuperviseddomainadaptationmethodbasedonrelativeentropyregularizationandmeasurepropagation
AT zhuopeng unsuperviseddomainadaptationmethodbasedonrelativeentropyregularizationandmeasurepropagation
AT yongjiasong unsuperviseddomainadaptationmethodbasedonrelativeentropyregularizationandmeasurepropagation
AT xiaoyiliu unsuperviseddomainadaptationmethodbasedonrelativeentropyregularizationandmeasurepropagation
AT huangqijiang unsuperviseddomainadaptationmethodbasedonrelativeentropyregularizationandmeasurepropagation
AT shubingliu unsuperviseddomainadaptationmethodbasedonrelativeentropyregularizationandmeasurepropagation
AT weixiwu unsuperviseddomainadaptationmethodbasedonrelativeentropyregularizationandmeasurepropagation
AT zhiyuanxiang unsuperviseddomainadaptationmethodbasedonrelativeentropyregularizationandmeasurepropagation