IDEAS home Printed from
   My bibliography  Save this paper

Wasserstein Dictionary Learning: Optimal Transport-based unsupervised non-linear dictionary learning


  • Morgan A. Schmitz

    (Astrophysics Department; IRFU; CEA; Université Paris-Saclay)

  • Matthieu Heitz

    (Université de Lyon; CNRS; LIRIS)

  • Nicolas Bonneel

    (Université de Lyon; CNRS; LIRIS)

  • Fred Ngolè

    (LIST, Data Analysis Tools Laboratory, CEA Saclay)

  • David Coeurjolly

    (Université de Lyon; CNRS; LIRIS)


This article introduces a new non-linear dictionary learning method for histograms in the probability simplex. The method leverages optimal transport theory, in the sense that our aim is to reconstruct histograms using so called displacement interpolations (a.k.a. Wasserstein barycenters) between dictionary atoms; such atoms are themselves synthetic histograms in the probability simplex. Our method simultaneously estimates such atoms, and, for each datapoint, the vector of weights that an optimally reconstruct it as an optimal transport barycenter of such atoms. Our method is computationally tractable thanks to the addition of an entropic regularization to the usual optimal transportation problem, leading to an approximation scheme that is e cient, parallel and simple to differentiate. Both atoms and weights are learned using a gradient-based descent method. Gradients are obtained by automatic di erentiation of the generalized Sinkhorn iterations that yield barycenters with entropic smoothing. Because of its formulation relying on Wasserstein barycenters instead of the usual matrix product between dictionary and codes, our method allows for non-linear relationships between atoms and the reconstruction of input data. We illustrate its application in several different image processing settings. ;Classification-JEL: 33F05, 49M99, 65D99, 90C08

Suggested Citation

  • Morgan A. Schmitz & Matthieu Heitz & Nicolas Bonneel & Fred Ngolè & David Coeurjolly, 2017. "Wasserstein Dictionary Learning: Optimal Transport-based unsupervised non-linear dictionary learning," Working Papers 2017-84, Center for Research in Economics and Statistics.
  • Handle: RePEc:crs:wpaper:2017-84

    Download full text from publisher

    File URL:
    File Function: CREST working paper version
    Download Restriction: no

    References listed on IDEAS

    1. Daniel D. Lee & H. Sebastian Seung, 1999. "Learning the parts of objects by non-negative matrix factorization," Nature, Nature, vol. 401(6755), pages 788-791, October.
    2. H. W. Kuhn, 1955. "The Hungarian method for the assignment problem," Naval Research Logistics Quarterly, John Wiley & Sons, vol. 2(1‐2), pages 83-97, March.
    3. Bassetti, Federico & Bodini, Antonella & Regazzini, Eugenio, 2006. "On minimum Kantorovich distance estimators," Statistics & Probability Letters, Elsevier, vol. 76(12), pages 1298-1302, July.
    Full references (including those not matched with items on IDEAS)

    Most related items

    These are the items that most often cite the same works as this one and are cited by the same works as this one.
    1. Mohammadamin Edrisi & Xiru Huang & Huw A. Ogilvie & Luay Nakhleh, 2023. "Accurate integration of single-cell DNA and RNA for analyzing intratumor heterogeneity using MaCroDNA," Nature Communications, Nature, vol. 14(1), pages 1-15, December.
    2. Del Corso, Gianna M. & Romani, Francesco, 2019. "Adaptive nonnegative matrix factorization and measure comparisons for recommender systems," Applied Mathematics and Computation, Elsevier, vol. 354(C), pages 164-179.
    3. P Fogel & C Geissler & P Cotte & G Luta, 2022. "Applying separative non-negative matrix factorization to extra-financial data," Working Papers hal-03689774, HAL.
    4. Xiao-Bai Li & Jialun Qin, 2017. "Anonymizing and Sharing Medical Text Records," Information Systems Research, INFORMS, vol. 28(2), pages 332-352, June.
    5. Weiqiang Shen & Chuanlin Zhang & Xiaona Zhang & Jinglun Shi, 2019. "A fully distributed deployment algorithm for underwater strong k-barrier coverage using mobile sensors," International Journal of Distributed Sensor Networks, , vol. 15(4), pages 15501477198, April.
    6. Naiyang Guan & Lei Wei & Zhigang Luo & Dacheng Tao, 2013. "Limited-Memory Fast Gradient Descent Method for Graph Regularized Nonnegative Matrix Factorization," PLOS ONE, Public Library of Science, vol. 8(10), pages 1-10, October.
    7. András Frank, 2005. "On Kuhn's Hungarian Method—A tribute from Hungary," Naval Research Logistics (NRL), John Wiley & Sons, vol. 52(1), pages 2-5, February.
    8. Spelta, A. & Pecora, N. & Rovira Kaltwasser, P., 2019. "Identifying Systemically Important Banks: A temporal approach for macroprudential policies," Journal of Policy Modeling, Elsevier, vol. 41(1), pages 197-218.
    9. M. Moghadam & K. Aminian & M. Asghari & M. Parnianpour, 2013. "How well do the muscular synergies extracted via non-negative matrix factorisation explain the variation of torque at shoulder joint?," Computer Methods in Biomechanics and Biomedical Engineering, Taylor & Francis Journals, vol. 16(3), pages 291-301.
    10. Amit Kumar & Anila Gupta, 2013. "Mehar’s methods for fuzzy assignment problems with restrictions," Fuzzy Information and Engineering, Springer, vol. 5(1), pages 27-44, March.
    11. Markovsky, Ivan & Niranjan, Mahesan, 2010. "Approximate low-rank factorization with structured factors," Computational Statistics & Data Analysis, Elsevier, vol. 54(12), pages 3411-3420, December.
    12. Paul Fogel & Yann Gaston-Mathé & Douglas Hawkins & Fajwel Fogel & George Luta & S. Stanley Young, 2016. "Applications of a Novel Clustering Approach Using Non-Negative Matrix Factorization to Environmental Research in Public Health," IJERPH, MDPI, vol. 13(5), pages 1-14, May.
    13. Le Thi Khanh Hien & Duy Nhat Phan & Nicolas Gillis, 2022. "Inertial alternating direction method of multipliers for non-convex non-smooth optimization," Computational Optimization and Applications, Springer, vol. 83(1), pages 247-285, September.
    14. Nisse, Nicolas & Salch, Alexandre & Weber, Valentin, 2023. "Recovery of disrupted airline operations using k-maximum matching in graphs," European Journal of Operational Research, Elsevier, vol. 309(3), pages 1061-1072.
    15. Chae, Bongsug (Kevin), 2018. "The Internet of Things (IoT): A Survey of Topics and Trends using Twitter Data and Topic Modeling," 22nd ITS Biennial Conference, Seoul 2018. Beyond the boundaries: Challenges for business, policy and society 190376, International Telecommunications Society (ITS).
    16. Parvin Ahmadi & Iman Gholampour & Mahmoud Tabandeh, 2018. "Cluster-based sparse topical coding for topic mining and document clustering," Advances in Data Analysis and Classification, Springer;German Classification Society - Gesellschaft für Klassifikation (GfKl);Japanese Classification Society (JCS);Classification and Data Analysis Group of the Italian Statistical Society (CLADAG);International Federation of Classification Societies (IFCS), vol. 12(3), pages 537-558, September.
    17. Jingfeng Guo & Chao Zheng & Shanshan Li & Yutong Jia & Bin Liu, 2022. "BiInfGCN: Bilateral Information Augmentation of Graph Convolutional Networks for Recommendation," Mathematics, MDPI, vol. 10(17), pages 1-16, August.
    18. Jianfei Cao & Han Yang & Jianshu Lv & Quanyuan Wu & Baolei Zhang, 2023. "Estimating Soil Salinity with Different Levels of Vegetation Cover by Using Hyperspectral and Non-Negative Matrix Factorization Algorithm," IJERPH, MDPI, vol. 20(4), pages 1-15, February.
    19. Wang, Ketong & Porter, Michael D., 2018. "Optimal Bayesian clustering using non-negative matrix factorization," Computational Statistics & Data Analysis, Elsevier, vol. 128(C), pages 395-411.
    20. Bachtenkirch, David & Bock, Stefan, 2022. "Finding efficient make-to-order production and batch delivery schedules," European Journal of Operational Research, Elsevier, vol. 297(1), pages 133-152.


    All material on this site has been provided by the respective publishers and authors. You can help correct errors and omissions. When requesting a correction, please mention this item's handle: RePEc:crs:wpaper:2017-84. See general information about how to correct material in RePEc.

    If you have authored this item and are not yet registered with RePEc, we encourage you to do it here. This allows to link your profile to this item. It also allows you to accept potential citations to this item that we are uncertain about.

    If CitEc recognized a bibliographic reference but did not link an item in RePEc to it, you can help with this form .

    If you know of missing items citing this one, you can help us creating those links by adding the relevant references in the same way as above, for each refering item. If you are a registered author of this item, you may also want to check the "citations" tab in your RePEc Author Service profile, as there may be some citations waiting for confirmation.

    For technical questions regarding this item, or to correct its authors, title, abstract, bibliographic or download information, contact: Secretariat General (email available below). General contact details of provider: .

    Please note that corrections may take a couple of weeks to filter through the various RePEc services.

    IDEAS is a RePEc service. RePEc uses bibliographic data supplied by the respective publishers.