IDEAS home Printed from https://ideas.repec.org/a/spr/coopap/v85y2023i1d10.1007_s10589-023-00458-3.html
   My bibliography  Save this article

Simple approximative algorithms for free-support Wasserstein barycenters

Author

Listed:
  • Johannes von Lindheim

    (Technische Universität Berlin)

Abstract

Computing Wasserstein barycenters of discrete measures has recently attracted considerable attention due to its wide variety of applications in data science. In general, this problem is NP-hard, calling for practical approximative algorithms. In this paper, we analyze a well-known simple framework for approximating Wasserstein- $${\varvec{p}}$$ p barycenters, where we mainly consider the most common case $${\varvec{p}}={\varvec{2}}$$ p = 2 and $${\varvec{p}}={\varvec{1}}$$ p = 1 , which is not as well discussed. The framework produces sparse support solutions and shows good numerical results in the free-support setting. Depending on the desired level of accuracy, this requires only $${\varvec{N}}-{\varvec{1}}$$ N - 1 or $${\varvec{N(N}}-{\varvec{1)/2 }}$$ N ( N - 1 ) / 2 standard two-marginal optimal transport (OT) computations between the $${\varvec{N}}$$ N input measures, respectively, which is fast, memory-efficient and easy to implement using any OT solver as a black box. What is more, these methods yield a relative error of at most $${\varvec{N}}$$ N and $${\varvec{2}}$$ 2 , respectively, for both $${\varvec{p}}={\varvec{ 1, 2}}$$ p = 1 , 2 . We show that these bounds are practically sharp. In light of the hardness of the problem, it is not surprising that such guarantees cannot be close to optimality in general. Nevertheless, these error bounds usually turn out to be drastically lower for a given particular problem in practice and can be evaluated with almost no computational overhead, in particular without knowledge of the optimal solution. In our numerical experiments, this guaranteed errors of at most a few percent.

Suggested Citation

  • Johannes von Lindheim, 2023. "Simple approximative algorithms for free-support Wasserstein barycenters," Computational Optimization and Applications, Springer, vol. 85(1), pages 213-246, May.
  • Handle: RePEc:spr:coopap:v:85:y:2023:i:1:d:10.1007_s10589-023-00458-3
    DOI: 10.1007/s10589-023-00458-3
    as

    Download full text from publisher

    File URL: http://link.springer.com/10.1007/s10589-023-00458-3
    File Function: Abstract
    Download Restriction: Access to the full text of the articles in this series is restricted.

    File URL: https://libkey.io/10.1007/s10589-023-00458-3?utm_source=ideas
    LibKey link: if access is restricted and if your library uses this service, LibKey will redirect you to where you can use your library subscription to access this item
    ---><---

    As the access to this document is restricted, you may want to search for a different version of it.

    References listed on IDEAS

    as
    1. G. Carlier & I. Ekeland, 2010. "Matching for teams," Economic Theory, Springer;Society for the Advancement of Economic Theory (SAET), vol. 42(2), pages 397-418, February.
    2. Ethan Anderes & Steffen Borgwardt & Jacob Miller, 2016. "Discrete Wasserstein barycenters: optimal transport for discrete data," Mathematical Methods of Operations Research, Springer;Gesellschaft für Operations Research (GOR);Nederlands Genootschap voor Besliskunde (NGB), vol. 84(2), pages 389-409, October.
    3. Puccetti, Giovanni & Rüschendorf, Ludger & Vanduffel, Steven, 2020. "On the computation of Wasserstein barycenters," Journal of Multivariate Analysis, Elsevier, vol. 176(C).
    4. repec:dau:papers:123456789/6728 is not listed on IDEAS
    5. Amir Beck & Shoham Sabach, 2015. "Weiszfeld’s Method: Old and New Results," Journal of Optimization Theory and Applications, Springer, vol. 164(1), pages 1-40, January.
    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. Steffen Borgwardt, 2022. "An LP-based, strongly-polynomial 2-approximation algorithm for sparse Wasserstein barycenters," Operational Research, Springer, vol. 22(2), pages 1511-1551, April.
    2. Florian Gunsilius & Meng Hsuan Hsieh & Myung Jin Lee, 2022. "Tangential Wasserstein Projections," Papers 2207.14727, arXiv.org, revised Aug 2022.
    3. Daniel Bartl & Samuel Drapeau & Jan Obloj & Johannes Wiesel, 2020. "Sensitivity analysis of Wasserstein distributionally robust optimization problems," Papers 2006.12022, arXiv.org, revised Nov 2021.
    4. Victor Chernozhukov & Alfred Galichon & Marc Henry & Brendan Pass, 2021. "Identification of Hedonic Equilibrium and Nonseparable Simultaneous Equations," Journal of Political Economy, University of Chicago Press, vol. 129(3), pages 842-870.
    5. Ethan Anderes & Steffen Borgwardt & Jacob Miller, 2016. "Discrete Wasserstein barycenters: optimal transport for discrete data," Mathematical Methods of Operations Research, Springer;Gesellschaft für Operations Research (GOR);Nederlands Genootschap voor Besliskunde (NGB), vol. 84(2), pages 389-409, October.
    6. Brendan Pass, 2019. "Interpolating between matching and hedonic pricing models," Economic Theory, Springer;Society for the Advancement of Economic Theory (SAET), vol. 67(2), pages 393-419, March.
    7. Simeon Reich & Truong Minh Tuyen, 2023. "The Generalized Fermat–Torricelli Problem in Hilbert Spaces," Journal of Optimization Theory and Applications, Springer, vol. 196(1), pages 78-97, January.
    8. Victor Chernozhukov & Alfred Galichon & Marc Henry & Brendan Pass, 2018. "Single Market Nonparametric Identification of Multi-Attribute Hedonic Equilibrium Models," SciencePo Working papers Main hal-01169655, HAL.
    9. Steffen Borgwardt & Stephan Patterson, 2021. "On the computational complexity of finding a sparse Wasserstein barycenter," Journal of Combinatorial Optimization, Springer, vol. 41(3), pages 736-761, April.
    10. G. Carlier & I. Ekeland, 2019. "Equilibrium in quality markets, beyond the transferable case," Economic Theory, Springer;Society for the Advancement of Economic Theory (SAET), vol. 67(2), pages 379-391, March.
    11. Victor Chernozhukov & Pierre-André Chiappori & Marc Henry, 2010. "Introduction," Economic Theory, Springer;Society for the Advancement of Economic Theory (SAET), vol. 42(2), pages 271-273, February.
    12. Steffen Borgwardt & Felix Happach, 2019. "Good Clusterings Have Large Volume," Operations Research, INFORMS, vol. 67(1), pages 215-231, January.
    13. Brendan Pass, 2017. "Interpolating between matching and hedonic pricing models," Papers 1701.04431, arXiv.org.
    14. Job Boerma & Aleh Tsyvinski & Alexander P. Zimin, 2021. "Sorting with Teams," Papers 2109.02730, arXiv.org, revised Nov 2023.
    15. Pierre-Andr'e Chiappori & Alfred Galichon & Bernard Salani'e, 2021. "On Human Capital and Team Stability," Papers 2102.06487, arXiv.org.
    16. Xinyang Wang, 2020. "Cooperation in Small Groups -- an Optimal Transport Approach," Papers 2005.11244, arXiv.org.
    17. Zhao, Yongning & Pan, Shiji & Zhao, Yuan & Liao, Haohan & Ye, Lin & Zheng, Yingying, 2024. "Ultra-short-term wind power forecasting based on personalized robust federated learning with spatial collaboration," Energy, Elsevier, vol. 288(C).
    18. Godichon-Baggioni, Antoine & Lu, Wei, 2024. "Online stochastic Newton methods for estimating the geometric median and applications," Journal of Multivariate Analysis, Elsevier, vol. 202(C).
    19. Puccetti, Giovanni & Rüschendorf, Ludger & Vanduffel, Steven, 2020. "On the computation of Wasserstein barycenters," Journal of Multivariate Analysis, Elsevier, vol. 176(C).
    20. Simone Görner & Christian Kanzow, 2016. "On Newton’s Method for the Fermat–Weber Location Problem," Journal of Optimization Theory and Applications, Springer, vol. 170(1), pages 107-118, July.

    Corrections

    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:spr:coopap:v:85:y:2023:i:1:d:10.1007_s10589-023-00458-3. 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: Sonal Shukla or Springer Nature Abstracting and Indexing (email available below). General contact details of provider: http://www.springer.com .

    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.