Full Text:   <5859>

CLC number: TP391

On-line Access: 2023-12-04

Received: 2022-12-09

Revision Accepted: 2023-12-05

Crosschecked: 2023-04-13

Cited: 0

Clicked: 953

Citations:  Bibtex RefMan EndNote GB/T7714

 ORCID:

Yahong Han

https://orcid.org/0000-0003-2768-1398

Runhua JIANG

https://orcid.org/0000-0003-2402-8684

-   Go to

Article info.
Open peer comments

Frontiers of Information Technology & Electronic Engineering  2023 Vol.24 No.11 P.1616-1632

http://doi.org/10.1631/FITEE.2200631


Dynamic parameterized learning for unsupervised domain adaptation


Author(s):  Runhua JIANG, Yahong HAN

Affiliation(s):  College of Intelligence and Computing, Tianjin University, Tianjin 300350, China; more

Corresponding email(s):   ddghjikle1@gmail.com, yahong@tju.edu.cn

Key Words:  Unsupervised domain adaptation, Optimization steps, Domain alignment, Semantic discrimination


Runhua JIANG, Yahong HAN. Dynamic parameterized learning for unsupervised domain adaptation[J]. Frontiers of Information Technology & Electronic Engineering, 2023, 24(11): 1616-1632.

@article{title="Dynamic parameterized learning for unsupervised domain adaptation",
author="Runhua JIANG, Yahong HAN",
journal="Frontiers of Information Technology & Electronic Engineering",
volume="24",
number="11",
pages="1616-1632",
year="2023",
publisher="Zhejiang University Press & Springer",
doi="10.1631/FITEE.2200631"
}

%0 Journal Article
%T Dynamic parameterized learning for unsupervised domain adaptation
%A Runhua JIANG
%A Yahong HAN
%J Frontiers of Information Technology & Electronic Engineering
%V 24
%N 11
%P 1616-1632
%@ 2095-9184
%D 2023
%I Zhejiang University Press & Springer
%DOI 10.1631/FITEE.2200631

TY - JOUR
T1 - Dynamic parameterized learning for unsupervised domain adaptation
A1 - Runhua JIANG
A1 - Yahong HAN
J0 - Frontiers of Information Technology & Electronic Engineering
VL - 24
IS - 11
SP - 1616
EP - 1632
%@ 2095-9184
Y1 - 2023
PB - Zhejiang University Press & Springer
ER -
DOI - 10.1631/FITEE.2200631


Abstract: 
unsupervised domain adaptation enables neural networks to transfer from a labeled source domain to an unlabeled target domain by learning domain-invariant representations. Recent approaches achieve this by directly matching the marginal distributions of these two domains. Most of them, however, ignore exploration of the dynamic trade-off between domain alignment and semantic discrimination learning, thus rendering them susceptible to the problems of negative transfer and outlier samples. To address these issues, we introduce the dynamic parameterized learning framework. First, by exploring domain-level semantic knowledge, the dynamic alignment parameter is proposed, to adaptively adjust the optimization steps of domain alignment and semantic discrimination learning. Besides, for obtaining semantic-discriminative and domain-invariant representations, we propose to align training trajectories on both source and target domains. Comprehensive experiments are conducted to validate the effectiveness of the proposed methods, and extensive comparisons are conducted on seven datasets of three visual tasks to demonstrate their practicability.

无监督域自适应的动态参数化学习

蒋润华1,2,韩亚洪1,2
1天津大学智能与计算学部,中国天津市,300350
2天津大学天津市机器学习重点实验室,中国天津市,300350
摘要:无监督领域自适应通过学习域不变表示实现神经网络从有标签数据组成的源域到无标签数据组成的目标域迁移。近期研究通过直接匹配这两个域的边缘分布实现这一目标。然而,已有研究大多数忽略域对齐和语义判别学习之间的动态平衡,因此容易受负迁移和异常样本影响。为解决这些问题,引入动态参数化学习框架。首先,通过探索领域级语义知识,提出动态对齐参数自适应地调整域对齐和语义判别学习的优化过程。此外,为获得判别能力强和域不变的表示,提出在源域和目标域上对齐优化过程。本文通过综合实验证明了所提出方法的有效性,并在3个视觉任务的7个数据集上进行广泛比较,证明可行性。

关键词:无监督领域自适应;优化步骤;跨域判别表示;语义判别

Darkslateblue:Affiliate; Royal Blue:Author; Turquoise:Article

Reference

[1]Ahmed SM, Raychaudhuri DS, Paul S, et al., 2021. Unsupervised multi-source domain adaptation without access to source data. Proc IEEE/CVF Conf on Computer Vision and Pattern Recognition, p.10103-10112.

[2]Bai Y, Wang C, Lou YH, et al., 2021. Hierarchical connectivity-centered clustering for unsupervised domain adaptation on person re-identification. IEEE Trans Image Process, 30:6715-6729.

[3]Bai ZC, Wang ZG, Wang J, et al., 2021. Unsupervised multi-source domain adaptation for person re-identification. Proc IEEE/CVF Conf on Computer Vision and Pattern Recognition, p.12914-12923.

[4]Balgi S, Dukkipati A, 2022. Contradistinguisher: a Vapnik’s imperative to unsupervised domain adaptation. IEEE Trans Patt Anal Mach Intell, 44(9):4730-4747.

[5]Cazenavette G, Wang TZ, Torralba A, et al., 2022. Dataset distillation by matching training trajectories. Proc IEEE/CVF Conf on Computer Vision and Pattern Recognition, p.4750-4759.

[6]Chang WG, You T, Seo S, et al., 2019. Domain-specific batch normalization for unsupervised domain adaptation. Proc IEEE/CVF Conf on Computer Vision and Pattern Recognition, p.7354-7362.

[7]Chen MH, Zhao S, Liu HF, et al., 2020. Adversarial-learned loss for domain adaptation. Proc 34th AAAI Conf on Artificial Intelligence, p.3521-3528.

[8]Chen XY, Wang SN, Wang JM, et al., 2021. Representation subspace distance for domain adaptation regression. Proc 38th Int Conf on Machine Learning, p.1749-1759.

[9]Courty N, Flamary R, Habrard A, et al., 2017. Joint distribution optimal transportation for domain adaptation. Proc 31st Int Conf on Neural Information Processing Systems, p.3733-3742.

[10]Dai YX, Liu J, Sun YF, et al., 2021. IDM: an intermediate domain module for domain adaptive person Re-ID. Proc IEEE/CVF Int Conf on Computer Vision, p.11864-11874.

[11]Fu Y, Wei YC, Wang GS, et al., 2019. Self-similarity grouping: a simple unsupervised cross domain adaptation approach for person re-identification. Proc IEEE/CVF Int Conf on Computer Vision, p.6112-6121.

[12]Ganin Y, Lempitsky V, 2015. Unsupervised domain adaptation by backpropagation. Proc 32nd Int Conf on Machine Learning, p.1180-1189.

[13]Ganin Y, Ustinova E, Ajakan H, et al., 2016. Domain-adversarial training of neural networks. In: Csurka G, (Ed.), Domain Adaptation in Computer Vision Applications. Advances in Computer Vision and Pattern Recognition. Springer, Cham, p.189-209.

[14]Ge Y, Chen DP, Li HS, 2020. Mutual mean-teaching: pseudo label refinery for unsupervised domain adaptation on person re-identification. Proc 8th Int Conf on Learning Representations.

[15]Gondal MW, Wuthrich M, Miladinovic D, et al., 2019. On the transfer of inductive bias from simulation to the real world: a new disentanglement dataset. Proc 33rd Int Conf on Neural Information Processing Systems, Article 1410.

[16]Han YH, Wu AM, Zhu LC, et al., 2021. Visual commonsense reasoning with directional visual connections. Front Inform Technol Electron Eng, 22(5):625-637.

[17]Han ZY, Sun HL, Yin YL, 2022. Learning transferable parameters for unsupervised domain adaptation. IEEE Trans Image Process, 31:6424-6439.

[18]He KM, Zhang XY, Ren SQ, et al., 2016. Deep residual learning for image recognition. Proc IEEE Conf on Computer Vision and Pattern Recognition, p.770-778.

[19]Higgins I, Matthey L, Pal A, et al., 2016. beta-VAE: learning basic visual concepts with a constrained variational framework. Proc 5th Int Conf on Learning Representations.

[20]Jing MM, Meng LC, Li JJ, et al., 2022. Adversarial mixup ratio confusion for unsupervised domain adaptation. IEEE Trans Multimedia, 25:2559-2572.

[21]Li MX, Zhai YM, Luo YW, et al., 2020. Enhanced transport distance for unsupervised domain adaptation. Proc IEEE/CVF Conf on Computer Vision and Pattern Recognition, p.13936-13944.

[22]Li QB, He BS, Song D, 2021. Model-contrastive federated learning. Proc IEEE/CVF Conf on Computer Vision and Pattern Recognitionn, p.10713-10722.

[23]Li S, Xie MX, Lv FR, et al., 2021a. Semantic concentration for domain adaptation. Proc IEEE/CVF Int Conf on Computer Vision, p.9102-9111.

[24]Li S, Xie MX, Gong KX, et al., 2021b. Transferable semantic augmentation for domain adaptation. Proc IEEE/CVF Conf on Computer Vision and Pattern Recognition, p.11516-11525.

[25]Li WK, Chen SC, 2022. Partial domain adaptation without domain alignment. IEEE Trans Patt Anal Mach Intell, 45(7):8787-8797.

[26]Li YJ, Lin CS, Lin YB, et al., 2019. Cross-dataset person re-identification via unsupervised pose disentanglement and adaptation. Proc IEEE/CVF Int Conf on Computer Vision, p.7919-7929.

[27]Li YY, Yao HT, Xu CS, 2022. Intra-domain consistency enhancement for unsupervised person re-identification. IEEE Trans Multimedia, 24:415-425.

[28]Liang J, Hu DP, Feng JS, 2020. Do we really need to access the source data? Source hypothesis transfer for unsupervised domain adaptation. Proc 37th Int Conf on Machine Learning, Article 560.

[29]Liu H, Wang JM, Long MS, 2021. Cycle self-training for domain adaptation. Proc 35th Advances in Neural Information Processing Systems, p.22968-22981.

[30]Long MS, Cao Y, Wang JM, et al., 2015. Learning transferable features with deep adaptation networks. Proc 32nd Int Conf on Machine Learning, p.97-105.

[31]Long MS, Zhu H, Wang JM, et al., 2017. Deep transfer learning with joint adaptation networks. Proc 34th Int Conf on Machine Learning, p.2208-2217.

[32]Lu YW, Li DS, Wang WJ, et al., 2021. Discriminative invariant alignment for unsupervised domain adaptation. IEEE Trans Multimedia, 24:1871-1882.

[33]Luo CC, Song CF, Zhang ZX, 2020. Generalizing person re-identification by camera-aware invariance learning and cross-domain mixup. Proc 16th European Conf on Computer Vision, p.224-241.

[34]Luo YW, Ren CX, 2021. Conditional bures metric for domain adaptation. Proc IEEE/CVF Conf on Computer Vision and Pattern Recognition, p.13989-13998.

[35]Pan SJ, Tsang IW, Kwok JT, et al., 2011. Domain adaptation via transfer component analysis. IEEE Trans Neur Netw, 22(2):199-210.

[36]Pan YW, Yao T, Li YH, et al., 2019. Transferrable prototypical networks for unsupervised domain adaptation. Proc IEEE/CVF Conf on Computer Vision and Pattern Recognition, p.2239-2247.

[37]Peng XC, Bai QX, Xia XD, et al., 2019. Moment matching for multi-source domain adaptation. Proc IEEE/CVF Int Conf on Computer Vision, p.1406-1415.

[38]Rakshit S, Banerjee B, Roig G, et al., 2019. Unsupervised multi-source domain adaptation driven by deep adversarial ensemble learning. Proc 41st German Conf on Pattern Recognition, p.485-498.

[39]Ristani E, Solera F, Zou R, et al., 2016. Performance measures and a data set for multi-target, multi-camera tracking. European Conf on Computer Vision, p.17-35.

[40]Saenko K, Kulis B, Fritz M, et al., 2010. Adapting visual category models to new domains. Proc 11th European Conf on Computer Vision, p.213-226.

[41]Saito K, Watanabe K, Ushiku Y, et al., 2018. Maximum classifier discrepancy for unsupervised domain adaptation. Proc IEEE/CVF Conf on Computer Vision and Pattern Recognition, p.3723-3732.

[42]Sun BC, Feng JS, Saenko K, 2016. Return of frustratingly easy domain adaptation. Proc 30th AAAI Conf on Artificial Intelligence, p.2058-2065.

[43]Tanwisuth K, Fan XJ, Zheng HJ, et al., 2021. A prototype-oriented framework for unsupervised domain adaptation. Proc 35th Conf on Neural Information Processing Systems, p.17194-17208.

[44]Tao XF, Kong J, Jiang M, et al., 2022. Unsupervised domain adaptation by multi-loss gap minimization learning for person re-identification. IEEE Trans Circ Syst Video Technol, 32(7):4404-4416.

[45]Tian Q, Zhu YN, Sun HY, et al., 2022. Unsupervised domain adaptation through dynamically aligning both the feature and label spaces. IEEE Trans Circ Syst Video Technol, 32(12):8562-8573.

[46]Venkat N, Kundu JN, Singh DK, et al., 2020. Your classifier can secretly suffice multi-source domain adaptation. Proc 34th Int Conf on Neural Information Processing Systems, Article 390.

[47]Venkateswara H, Eusebio J, Chakraborty S, et al., 2017. Deep hashing network for unsupervised domain adaptation. Proc IEEE Conf on Computer Vision and Pattern Recognition, p.5018-5027.

[48]Wang DK, Zhang SL, 2020. Unsupervised person re-identification via multi-label classification. Proc IEEE/CVF Conf on Computer Vision and Pattern Recognition, p.10981-10990.

[49]Wang W, Zhao F, Liao S, et al., 2022. Attentive waveblock: complementarity-enhanced mutual networks for unsupervised domain adaptation in person re-identification and beyond. IEEE Trans Image Process, 31:1532-1544.

[50]Wang XM, Li L, Ye WR, et al., 2019. Transferable attention for domain adaptation. Proc 33rd AAAI Conf on Artificial Intelligence, p.5345-5352.

[51]Wei GQ, Lan CL, Zeng WJ, et al., 2021. MetaAlign: coordinating domain alignment and classification for unsupervised domain adaptation. Proc IEEE/CVF Conf on Computer Vision and Pattern Recognition, p.16643-16653.

[52]Wei LH, Zhang SL, Gao W, et al., 2018. Person transfer GAN to bridge domain gap for person re-identification. Proc IEEE/CVF Conf on Computer Vision and Pattern Recognition, p.79-88.

[53]Wu AM, Han YH, Zhu LC, et al., 2022. Instance-invariant domain adaptive object detection via progressive disentanglement. IEEE Trans Patt Anal Mach Intell, 44(8):4178-4193.

[54]Wu KH, Jia F, Han YH, 2023. Domain-specific feature elimination: multi-source domain adaptation for image classification. Front Comput Sci, 17(4):174705.

[55]Xiao N, Zhang L, 2021. Dynamic weighted learning for unsupervised domain adaptation. Proc IEEE/CVF Conf on Computer Vision and Pattern Recognition, p.15242-15251.

[56]Xu RJ, Chen ZL, Zuo WM, et al., 2018. Deep cocktail network: multi-source unsupervised domain adaptation with category shift. Proc IEEE/CVF Conf on Computer Vision and Pattern Recognition, p.3964-3973.

[57]Xu RJ, Li GB, Yang JH, et al., 2019. Larger norm more transferable: an adaptive feature norm approach for unsupervised domain adaptation. Proc IEEE/CVF Int Conf on Computer Vision, p.1426-1435.

[58]Xu X, Zhang LY, 2021. Rectifying pseudo label by mutual disagreement learning for unsupervised domain adaptation person re-identification. Proc IEEE Int Conf on Multimedia and Expo, p.1-6.

[59]Xu YY, Yan H, 2022. Cycle-reconstructive subspace learning with class discriminability for unsupervised domain adaptation. Patt Recognit, 129:108700.

[60]Yang SQ, van de Weijer J, Herranz L, et al., 2021. Exploiting the intrinsic neighborhood structure for source-free domain adaptation. Proc 35th Advances in Neural Information Processing Systems, p.29393-29405.

[61]Yang YC, Soatto S, 2020. FDA: Fourier domain adaptation for semantic segmentation. Proc IEEE/CVF Conf on Computer Vision and Pattern Recognition, p.4085-4095.

[62]Zhai YP, Lu SJ, Ye QX, et al., 2020. AD-Cluster: augmented discriminative clustering for domain adaptive person re-identification. Proc IEEE/CVF Conf on Computer Vision and Pattern Recognition, p.9021-9030.

[63]Zhang H, Cao HH, Yang X, et al., 2021. Self-training with progressive representation enhancement for unsupervised cross-domain person re-identification. IEEE Trans Image Process, 30:5287-5298.

[64]Zhang JY, Huang JX, Tian ZC, et al., 2022. Spectral unsupervised domain adaptation for visual recognition. Proc IEEE/CVF Conf on Computer Vision and Pattern Recognition, p.9829-9840.

[65]Zhang MY, Liu K, Li YD, et al., 2021. Unsupervised domain adaptation for person re-identification via heterogeneous graph alignment. Proc 35th AAAI Conf on Artificial Intelligence, p.3360-3368.

[66]Zhang YB, Tang H, Jia K, et al., 2019. Domain-symmetric networks for adversarial domain adaptation. Proc IEEE/CVF Conf on Computer Vision and Pattern Recognition, p.5031-5040.

[67]Zhang YB, Deng B, Tang H, et al., 2020. Unsupervised multi-class domain adaptation: theory, algorithms, and practice. IEEE Trans Patt Anal Mach Intell, 44(5):2775-2792.

[68]Zhang YC, Liu TL, Long MS, et al., 2019. Bridging theory and algorithm for domain adaptation. Proc 36th Int Conf on Machine Learning, p.7404-7413.

[69]Zhang Z, Wang YN, Liu S, et al., 2022. Cross-domain person re-identification using heterogeneous convolutional network. IEEE Trans Circ Syst Video Technol, 32(3):1160-1171.

[70]Zhao F, Liao SC, Xie GS, et al., 2020. Unsupervised domain adaptation with noise resistible mutual-training for person re-identification. Proc 16th European Conf on Computer Vision, p.526-544.

[71]Zhao SC, Li B, Xu PF, et al., 2021. MADAN: multi-source adversarial domain aggregation network for domain adaptation. Int J Comput Vis, 129(8):2399-2424.

[72]Zheng L, Shen LY, Tian L, et al., 2015. Scalable person re-identification: a benchmark. Proc IEEE Int Conf on Computer Vision, p.1116-1124.

[73]Zhong Z, Zheng L, Luo ZM, et al., 2019. Invariance matters: exemplar memory for domain adaptive person re-identification. Proc IEEE/CVF Conf on Computer Vision and Pattern Recognition, p.598-607.

[74]Zhong Z, Zheng L, Luo ZM, et al., 2021. Learning to adapt invariance in memory for person re-identification. IEEE Trans Patt Anal Mach Intell, 43(8):2723-2738.

[75]Zhu YC, Zhuang FZ, Wang DQ, 2019. Aligning domain-specific distribution and classifier for cross-domain classification from multiple sources. Proc 33rd AAAI Conf on Artificial Intelligence, p.5989-5996.

[76]Zhu YC, Zhuang FZ, Wang JD, et al., 2021. Deep sub-domain adaptation network for image classification. IEEE Trans Neur Netw Learn Syst, 32(4):1713-1722.

[77]Zuo YK, Yao HT, Zhuang LS, et al., 2022. Margin-based adversarial joint alignment domain adaptation. IEEE Trans Circ Syst Video Technol, 32(4):2057-2067.

Open peer comments: Debate/Discuss/Question/Opinion

<1>

Please provide your name, email address and a comment





Journal of Zhejiang University-SCIENCE, 38 Zheda Road, Hangzhou 310027, China
Tel: +86-571-87952783; E-mail: cjzhang@zju.edu.cn
Copyright © 2000 - 2024 Journal of Zhejiang University-SCIENCE