Skip to main content
SpringerLink
Log in

Instance reweighting and dynamic distribution alignment for domain adaptation

  • Original Research
  • Published:
Journal of Ambient Intelligence and Humanized Computing Aims and scope Submit manuscript

Abstract

Domain adaptation aims to learn a robust classifier from source data that performs well on different target data with distinct distributions. This paper proposes a novel unsupervised domain adaptation method for real-world visual recognition, object recognition, and handwritten digit recognition tasks. Although previous domain adaptation methods had explored the properties of domains and attempted to align the domains, most of them have essential significant challenges. Some researches were just learning appropriate features, and some others were just reweighing the instances to mitigate the difference between domain distributions; however, both are essential in significant and complex datasets. To overcome this challenge, we propose a unified framework that considers the feature extraction approach and the instance reweighting approach, simultaneously. Moreover, the previous methods had ignored the different importance of the marginal and conditional distributions in the distribution alignment of the source and the target domains. However, this problem must be considered, which leads to better performance in the classification problems, especially when domains are dissimilar. Therefore, we also employ dynamic weighted marginal and conditional distributions according to their importance. The final problem is the absence of a confident adaptive classification technique. Therefore, our last proposal learns an adaptive domain-invariant classifier by structural risk minimization to obtain better classification results. The experimental results validate that our hybrid domain adaptation method significantly improves the classification accuracy compared to the state-of-the-art domain adaptation methods; in particular, experiments demonstrate that our proposed technique is even more powerful than the methods of domain adaptation using deep neural networks.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Log in via an institution

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Fig. 1

source domain; (b) similar domains; (c) different domains (Wang et al. 2018)

Fig. 2

source and the target domain data after joint feature matching and instance reweighting in the newly learned subspace. The irrelevant source instances are now down-weighted (the unfilled shapes) to reduce further the domain difference, (c) the classified source and target data

Fig. 3

source data, (b) unlabeled target data, (c) the source and the target domain data in the new spaces. The irrelevant source instances are now vanished (the calculator and the cellphone) to reduce further the domain difference, (d) the classified source and target data

Fig. 4
Fig. 5
Fig. 6
Fig. 7
Fig. 8
Fig. 9
Fig. 10
Fig. 11
Fig. 12
Fig. 13
Fig. 14
Fig. 15
Fig. 16

source domain (δ = 0), (b) without the instance reweighting (σ = 0), (c) ignoring the target data scatter (μ = 0), (d) disregarding the difference between projection matrices (λ = 0)

Similar content being viewed by others

References

  • Azarkesht M, Afsari F (2019a) Domain-invariant adaptation exploiting statistical and geometrical properties. In: 7th Iranian joint congress on fuzzy and intelligent systems (CFIS). https://doi.org/10.1109/CFIS.2019.8692162

  • Azarkesht M, Afsari F (2019b) Joint feature extraction and instance re-weighting approach for unsupervised domain adaptation. In: Conference on Contemporary issues in computer information and sciences

  • Baktashmotlagh M, Harandi MT, Lovell BC, Salzmann M (2013) Unsupervised domain adaptation by domain invariant projection. In: Proceedings of the IEEE International Conference on computer vision, pp 769–776

  • Bay H, Ess A, Tuytelaars T, Van Gool L (2008) Speeded-up robust features (SURF). Comput Vis Image Underst 110(3):346–359

    Article  Google Scholar 

  • Belkin M, Niyogi P (2002) Laplacian eigenmaps and spectral techniques for embedding and clustering. In: Advances in neural information processing systems, pp 585–591

  • Belkin M, Niyogi P, Sindhwani V (2015) Manifold regularization: a geometric framework for learning from labeled and unlabeled examples. J Mach Learn Res 7:2399–2434

    MathSciNet  MATH  Google Scholar 

  • Bruzzone L, Marconcini M (2010) Domain adaptation problems: a DASVM classification technique and a circular validation strategy. IEEE Trans Pattern Anal Mach Intell 32(5):770–787

    Article  Google Scholar 

  • Cao Y, Long M, Wang J (2018) Unsupervised domain adaptation with distribution matching machines. In: AAAI Conference on artificial intelligence

  • Chen M, Weinberger KQ, Blitzer JC (2011) Co-training for domain adaptation. In: Twenty-fifth Conference on Neural information processing systems, pp 2456–2464

  • Chen M, Weinberger KQ, Sha F, Angeles L (2012) Marginalized denoising autoencoders for domain adaptation. In: Proceedings of the 29th International Conference on Machine Learning (ICML-12), pp 767–774

  • Chen L, Zhang H, Xiao J, Liu W, Chang SF (2018) Zero-shot visual recognition using semantics-preserving adversarial embedding networks. In: Proceedings of the IEEE Conference on Computer vision and pattern recognition, pp 1043–1052

  • Chu WS, Torre FD La, Cohn JF (2013) Selective transfer machine for personalized facial action unit detection. In: Proceedings of the IEEE Computer society conference on computer vision and pattern recognition, pp 3515–3522

  • Csurka G (ed) (2017) Domain adaptation in computer vision applications. Springer international publishing, Cham

    Google Scholar 

  • Ding Z, Fu Y (2016) Robust transfer metric learning for image classification. IEEE Trans Image Process 26(2):660–670

    Article  MathSciNet  Google Scholar 

  • Ding Z, Fu Y (2018) Deep transfer low-rank coding for cross-domain learning. IEEE Trans Neural Netw Learn Syst 30(6):1768–1779

    Article  MathSciNet  Google Scholar 

  • Donahue J, Jia Y, Vinyals O, Hoffman J, Zhang N, Tzeng E, Darrell T (2014) Decaf: A deep convolutional activation feature for generic visual recognition. In: International Conference on Machine Learning, pp 647–655

  • Dou Q, Ouyang C, Chen C, Chen H, Heng PA (2018) Unsupervised cross-modality domain adaptation of convnets for biomedical image segmentations with adversarial loss. In: International joint conference on artificial intelligence, pp 691–697

  • Fernando B, Habrard A, Sebban M, Tuytelaars T, Leuven KU, Hubert L ,Lauras B (2013) Unsupervised visual domain adaptation using subspace alignment. In: Proceedings of the IEEE International conference on computer vision, pp 2960–2967

  • Ghifary M, Balduzzi D, Kleijn WB, Zhang M (2017) Scatter component analysis: a unified framework for domain adaptation and domain generalization. IEEE Trans Pattern Anal Mach Intell 39:1414–1430

    Article  Google Scholar 

  • Glorot X, Bordes A, Bengio Y (2011) Domain adaptation for large-scale sentiment classification: a deep learning approach. In: The 28th International conference on machine learning, pp 513–520

  • Gong B, Shi Y, Sha F, Grauman K (2012) Geodesic flow kernel for unsupervised domain adaptation. In: 2012 IEEE conference on computer vision and pattern recognition, pp 2066–2073

  • Gopalan R, Li R, Chellappa R (2011) Domain adaptation for object recognition: An unsupervised approach. In: 2011 international conference on computer vision, pp 999–1006

  • Griffin G, Holub A, Perona P (2007) Caltech-256 object category dataset. California institute of technology, California

    Google Scholar 

  • Krizhevsky A, Sutskever I, Hinton GE (2012) Imagenet classification with deep convolutional neural networks. Adv Neural Inf Process Syst 25:1097–1105

    Google Scholar 

  • Li S, Song S, Huang G, Ding Z, Wu C (2018) Domain invariant and class discriminative feature learning for visual domain adaptation. IEEE Trans Image Process 27(9):4260–4273. https://doi.org/10.1109/TIP.2018.2839528

    Article  MathSciNet  Google Scholar 

  • Long M, Ding G, Wang J, Sun J, Guo Y, Yu PS (2013a) Transfer sparse coding for robust image representation. In: Proceedings of the IEEE conference on computer vision and pattern recognition, pp 407–414. https://doi.org/10.1109/CVPR.2013.59

  • Long M, Wang J, Ding G, Pan SJ, Philip SY (2013b) Adaptation regularization: a general framework for transfer learning. IEEE Trans Knowl Data Eng 26(5):1076–1089

    Article  Google Scholar 

  • Long M, Wang J, Ding G, Sun J, Yu PS (2013c) Transfer feature learning with joint distribution adaptation. In: Proceedings of the IEEE international conference on computer vision, pp 2200–2207. https://doi.org/10.1109/ICCV.2013.274

  • Long M, Wang J, Ding G, Sun J, Yu PS (2014) Transfer joint matching for unsupervised domain adaptation. In: Proceedings of the IEEE conference on computer vision and pattern recognition, pp 1410–1417

  • Long M, Wang J, Sun J, Yu PS (2015) Domain invariant transfer kernel learning. IEEE Trans Knowl Data Eng. https://doi.org/10.1109/TKDE.2014.2373376

    Article  Google Scholar 

  • Masaeli M, Fung G, Dy JG (2010) From transformation-based dimensionality reduction to feature selection. In: Proceedings of the 27th international conference on machine learning (ICML-10), pp 751–758

  • Pan SJ, Yang Q (2010) A Survey on transfer learning. IEEE Trans Knowl Data Eng 22:1345–1359

    Article  Google Scholar 

  • Pan SJ, Ni X, Sun JT, Yang Q, Chen Z (2010) Cross-domain sentiment classification via spectral feature alignment. In: Proceedings of the 19th international conference on World wide web, pp 751–760. https://doi.org/10.1145/1772690.1772767

  • Pan SJ, Tsang IW, Kwok JT, Yang Q (2011) Domain adaptation via transfer component analysis. IEEE Trans Neural Netw Learn Syst 22(2):199–210

    Article  Google Scholar 

  • Patel VM, Gopalan R, Li R, Chellappa R (2015) Visual domain adaptation: a survey of recent advances. IEEE Signal Process Mag 32(3):53–69. https://doi.org/10.1109/MSP.2014.2347059

    Article  Google Scholar 

  • Saenko K, Kulis B, Fritz M, Darrell T (2010) Adapting visual category models to new domains, In: European Conference on Computer Vision, pp 213–226

  • Shlens J (2014) A tutorial on principal component analysis. In: arXiv preprint arXiv:1404.1100

  • Si S, Tao D, Geng B (2010) Bregman divergence-based regularization for transfer subspace learning. IEEE Trans Knowl Data Eng 22:929–942

    Article  Google Scholar 

  • Sun B, Saenko K (2015) Subspace distribution alignment for unsupervised domain adaptation, In: Proceedings of the British machine vision conference, pp 24–33

  • Sun B, Saenko K (2016) Deep coral: Correlation alignment for deep domain adaptation. In: European conference on computer vision, Springer, pp 443-450

  • Thopalli K, Anirudh R, Thiagarajan JJ, Turaga P (2018) Multiple subspace alignment improves domain adaptation. In: IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pp 3552–3556

  • Tzeng E, Hoffman J, Zhang N, Saenko K, Darrell T (2014) Deep domain confusion: Maximizing for domain invariance, In: arXiv preprint arXiv: pp 1412.3474. https://doi.org/10.1016/S0140-6736(83)90938-8

  • Valindria VV, Lavdas I, Bai W, Kamnitsas K, Aboagye EO, Rockall AG, Rueckert D, Glocker B (2018) Domain adaptation for MRI organ segmentation using reverse classification accuracy. arXiv preprint arXiv:1806.00363 ,pp 1–9

  • Wang C, Mahadevan S (2009) Manifold alignment without correspondence, In: Proceedings of the twenty-first international joint conference on artificial intelligence organization, pp 1273–1278

  • Wang H, Wang W, Zhang C, Xu F (2014) Cross-domain metric learning based on information theory. Proc AAAI Conf Artif Intell 28(1):2099–2105. https://doi.org/10.1143/JJAP.10.951

    Article  Google Scholar 

  • Wang J, Feng W, Chen Y, Yu H, Huang M, Yu PS (2018) Visual domain adaptation with manifold embedded distribution alignment, In: Proceedings of the 26th ACM international conference on Multimedia, pp 402–410. https://doi.org/10.1145/3240508.3240512

  • Xu Y, Fang X, Wu J, Li X, Zhang D (2016) Discriminative transfer subspace learning via low-rank and sparse representation. IEEE Trans Image Process. https://doi.org/10.1109/TIP.2015.2510498

    Article  MathSciNet  MATH  Google Scholar 

  • Yang H, Zhou X, Lin L, Yao B, Tan Z, Tang H, Tian Y (2018) Deep sparse informative transfer softmax for cross-domain image classification. In: International conference on database systems for advanced applications. Springer. Cham, pp 19–39. https://doi.org/10.1007/978-3-319-91458-9

  • Zhang J, Li W, Ogunbona P (2017) Joint geometrical and statistical alignment for visual domain adaptation. In: Proceeding of 30th IEEE Conference on Computer Vision and Pattern Recognition, pp 5150–5158. https://doi.org/10.1109/CVPR.2017.547

  • Zhang J, Li W, Ogunbona P (2019) Unsupervised domain adaptation: A multi-task learning-based method, Knowl-Based Syst, pp 10975. https://doi.org/10.1007/978-3-319-95333-5_14

  • Zhou D, Bousquet O, Lal TN, Weston J, Schölkopf B (2004) Learning with local and global consistency. In: Advances in neural information processing systems, pp 321–328

  • Zhuang F, Luo P, Shen Z, He Q, Xiong Y, Shi Z, Xiong H (2011) Mining distinction and commonality across multiple domains using generative model for text classification. IEEE Trans Knowl Data Eng 24:2025–2039

    Article  Google Scholar 

  • Zhuo J, Wang S, Zhang W, Huang Q (2017) Deep unsupervised convolutional domain adaptation. InProceedings of the 25th ACM international conference on Multimedia. In: Proceedings of the 2017 ACM on Multimedia conference-MM’17, pp 261–269. https://doi.org/10.1145/3123266.3123292

Download references

Author information

Authors and Affiliations

  1. Department of Computer Engineering, Faculty of Engineering, Shahid Bahonar University of Kerman, Afzalipoor square, PO Box 76169-133, Kerman, Iran

    Maryam Azarkesht & Fatemeh Afsari

Authors
  1. Maryam Azarkesht
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Fatemeh Afsari
    View author publications

    You can also search for this author in PubMed Google Scholar

Corresponding author

Correspondence to Fatemeh Afsari.

Ethics declarations

Conflict of interest

Author Maryam Azarkesht declares that she has no conflict of interest. Author Fatemeh Afsari declares that she has no conflict of interest.

Ethical approval

This article does not contain any studies with human participants or animals performed by any of the authors.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Azarkesht, M., Afsari, F. Instance reweighting and dynamic distribution alignment for domain adaptation. J Ambient Intell Human Comput 13, 4967–4987 (2022). https://doi.org/10.1007/s12652-021-03426-z

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s12652-021-03426-z

Keywords

Search

Navigation