Energy-based open set domain adaptation with dynamic weighted synergistic mechanism

Abstract Open Set Domain Adaptation (OSDA) aims to minimize domain variation while distinguishing between known and unknown samples. However, existing OSDA methods, which rely on deep neural network classifiers, often lead to overconfident predictions and fail to clearly demarcate known from unknown...

Full description

Saved in:
Bibliographic Details
Main Authors: Zihao Fu, Dong Liu, Shengsheng Wang, Hao Chai
Format: Article
Language:English
Published: Springer 2025-04-01
Series:Complex & Intelligent Systems
Subjects:
Online Access:https://doi.org/10.1007/s40747-025-01857-1
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850154652310437888
author Zihao Fu
Dong Liu
Shengsheng Wang
Hao Chai
author_facet Zihao Fu
Dong Liu
Shengsheng Wang
Hao Chai
author_sort Zihao Fu
collection DOAJ
description Abstract Open Set Domain Adaptation (OSDA) aims to minimize domain variation while distinguishing between known and unknown samples. However, existing OSDA methods, which rely on deep neural network classifiers, often lead to overconfident predictions and fail to clearly demarcate known from unknown samples. To address this limitation, we propose the Energy-based Open Set domain adaptation (EOS) method. EOS introduces a novel two-stage approach involving a separation stage followed by an alignment stage. In the separation stage, we employ an energy-based anomaly detection strategy to identify unknown samples, transforming the traditional K-class classification task into a K+1-dimensional classifier by introducing an additional dimension to model the uncertainty of out-of-distribution samples. To further refine separation, we apply a coarse-to-fine method that iteratively improves the separation outcomes, which are integrated as weighted inputs in the alignment process to enhance feature distribution alignment. In the alignment stage, we employ a dynamic weighted synergistic mechanism, where the separation network and alignment network co-evolve through continuous alternating training. This mechanism enables the system to better adapt to invariant features across domains. We evaluate EOS on standard benchmarks, including Office-31, Office-Home, and VisDA-2017, with experimental results demonstrating that EOS consistently outperforms other state-of-the-art methods.
format Article
id doaj-art-65b5e6855aa24f40a9e3a8546ee67a5d
institution OA Journals
issn 2199-4536
2198-6053
language English
publishDate 2025-04-01
publisher Springer
record_format Article
series Complex & Intelligent Systems
spelling doaj-art-65b5e6855aa24f40a9e3a8546ee67a5d2025-08-20T02:25:16ZengSpringerComplex & Intelligent Systems2199-45362198-60532025-04-0111611310.1007/s40747-025-01857-1Energy-based open set domain adaptation with dynamic weighted synergistic mechanismZihao Fu0Dong Liu1Shengsheng Wang2Hao Chai3Hunan Engineering Research Center of Advanced Embedded Computing and Intelligent Medical Systems, Xiangnan UniversityHunan Engineering Research Center of Advanced Embedded Computing and Intelligent Medical Systems, Xiangnan UniversityCollege of Computer Science and Technology, Jilin UniversityHunan Engineering Research Center of Advanced Embedded Computing and Intelligent Medical Systems, Xiangnan UniversityAbstract Open Set Domain Adaptation (OSDA) aims to minimize domain variation while distinguishing between known and unknown samples. However, existing OSDA methods, which rely on deep neural network classifiers, often lead to overconfident predictions and fail to clearly demarcate known from unknown samples. To address this limitation, we propose the Energy-based Open Set domain adaptation (EOS) method. EOS introduces a novel two-stage approach involving a separation stage followed by an alignment stage. In the separation stage, we employ an energy-based anomaly detection strategy to identify unknown samples, transforming the traditional K-class classification task into a K+1-dimensional classifier by introducing an additional dimension to model the uncertainty of out-of-distribution samples. To further refine separation, we apply a coarse-to-fine method that iteratively improves the separation outcomes, which are integrated as weighted inputs in the alignment process to enhance feature distribution alignment. In the alignment stage, we employ a dynamic weighted synergistic mechanism, where the separation network and alignment network co-evolve through continuous alternating training. This mechanism enables the system to better adapt to invariant features across domains. We evaluate EOS on standard benchmarks, including Office-31, Office-Home, and VisDA-2017, with experimental results demonstrating that EOS consistently outperforms other state-of-the-art methods.https://doi.org/10.1007/s40747-025-01857-1Domain adaptationOpen set domain adaptationOut-of-distribution detectionEnergy-based models
spellingShingle Zihao Fu
Dong Liu
Shengsheng Wang
Hao Chai
Energy-based open set domain adaptation with dynamic weighted synergistic mechanism
Complex & Intelligent Systems
Domain adaptation
Open set domain adaptation
Out-of-distribution detection
Energy-based models
title Energy-based open set domain adaptation with dynamic weighted synergistic mechanism
title_full Energy-based open set domain adaptation with dynamic weighted synergistic mechanism
title_fullStr Energy-based open set domain adaptation with dynamic weighted synergistic mechanism
title_full_unstemmed Energy-based open set domain adaptation with dynamic weighted synergistic mechanism
title_short Energy-based open set domain adaptation with dynamic weighted synergistic mechanism
title_sort energy based open set domain adaptation with dynamic weighted synergistic mechanism
topic Domain adaptation
Open set domain adaptation
Out-of-distribution detection
Energy-based models
url https://doi.org/10.1007/s40747-025-01857-1
work_keys_str_mv AT zihaofu energybasedopensetdomainadaptationwithdynamicweightedsynergisticmechanism
AT dongliu energybasedopensetdomainadaptationwithdynamicweightedsynergisticmechanism
AT shengshengwang energybasedopensetdomainadaptationwithdynamicweightedsynergisticmechanism
AT haochai energybasedopensetdomainadaptationwithdynamicweightedsynergisticmechanism