kth.sePublications
Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
On the complexity of the optimal transport problem with graph-structured cost
Georgia Tech, Atlanta, GA 30332 USA.
KTH, School of Engineering Sciences (SCI), Mathematics (Dept.).ORCID iD: 0000-0002-2484-0181
KTH, School of Engineering Sciences (SCI), Mathematics (Dept.), Optimization and Systems Theory.ORCID iD: 0000-0001-5158-9255
Georgia Tech, Atlanta, GA 30332 USA.
2022 (English)In: Proceedings of The 25th International Conference on Artificial Intelligence and Statistics, PMLR , 2022, Vol. 151, p. 9147-9165Conference paper, Published paper (Refereed)
Abstract [en]

Multi-marginal optimal transport (MOT) is a generalization of optimal transport to multiple marginals. Optimal transport has evolved into an important tool in many machine learning applications, and its multi-marginal extension opens up for addressing new challenges in the field of machine learning. However, the usage of MOT has been largely impeded by its computational complexity which scales exponentially in the number of marginals. Fortunately, in many applications, such as barycenter or interpolation problems, the cost function adheres to structures, which has recently been exploited for developing efficient computational methods. In this work we derive computational bounds for these methods. In particular, with $m$ marginal distributions supported on $n$ points, we provide a $ \mathcal\tilde O(d(\mathcalT)m n^w(G)+1ε^-2)$ bound for a ε-accuracy when the problem is associated with a graph that can be factored as a junction tree with diameter $d(\mathcalT)$ and tree-width $w(G)$. For the special case of the Wasserstein barycenter problem, which corresponds to a star-shaped tree, our bound is in alignment with the existing complexity bound for it.

Place, publisher, year, edition, pages
PMLR , 2022. Vol. 151, p. 9147-9165
Series
Proceedings of Machine Learning Research
National Category
Computational Mathematics
Identifiers
URN: urn:nbn:se:kth:diva-312088ISI: 000841852303024OAI: oai:DiVA.org:kth-312088DiVA, id: diva2:1657456
Conference
The 25th International Conference on Artificial Intelligence and Statistics, MAR 28-30, 2022, ELECTR NETWORK
Note

QC 20220511

Available from: 2022-05-11 Created: 2022-05-11 Last updated: 2022-11-14Bibliographically approved
In thesis
1. Graph-structured multi-marginal optimal transport: Theory, applications, and efficient methods using entropy regularization
Open this publication in new window or tab >>Graph-structured multi-marginal optimal transport: Theory, applications, and efficient methods using entropy regularization
2022 (English)Doctoral thesis, comprehensive summary (Other academic)
Abstract [en]

This thesis deals with a class of multi-marginal optimal transport problems, which we call graph-structured multi-marginal optimal transport. The aim of the thesis is to work towards a unified framework for this class of problems. The included papers explore theoretical, computational, and practical aspects of the novel framework, e.g., connections to related areas, numerical algorithms for solving the problem, and applications in various fields. The first paper treats multi-marginal optimal transport problems with cost functions that decouple according to a tree. We show that the entropy regularized version of this problem can be seen as a generalization of the Schrödinger bridge problem to the same underlying tree. This connection is utilized to derive an efficient method for approximately solving the optimal transport problem. The second paper applies the framework to inverse problems in radar imaging. In this setting only partial information of the marginals in the optimal transport problem are available. We extend the methods from the first paper to handle the case of partial information. Several numerical examples illustrate that the framework can be used to find robust estimates of spatial spectra in information fusion and tracking applications. In the third paper we model network flow problems by utilizing graphstructured multi-marginal optimal transport. In particular, we show that the dynamic minimum-cost multi-commodity network flow problem can be formulated as a multi-marginal optimal transport problem that has an underlying structure, which can be described by a graph that contains cycles. The computational methods from the previous papers are extended to efficiently solve network flow problems. The fourth paper applies the graph-structured multi-marginal optimal transport framework to estimation and control problems for multi-agent systems. Moreover, the framework is utilized to prove a duality result between estimation and control of multi-agent systems. The fifth paper explores connections between graph-structured multimarginal optimal transport and probabilistic graphical models. We show that the entropy regularized optimal transport problem is equivalent to an inference problem for probabilistic graphical models with constraints on some of the marginal distributions. The sixth paper proves complexity results for graph-structured multimarginal optimal transport. Moreover, based on a junction tree factorization of the underlying graph we provide a general algorithm for approximately solving graph-structured multi-marginal optimal transport. Our complexity bounds match the best known results for the optimal transport barycenter problem and the general multi-marginal optimal transport problem.

Place, publisher, year, edition, pages
Stockholm: KTH Royal Institute of Technology, 2022. p. 307
Series
TRITA-SCI-FOU ; 2022:22
National Category
Control Engineering Computational Mathematics
Research subject
Applied and Computational Mathematics, Optimization and Systems Theory
Identifiers
urn:nbn:se:kth:diva-312138 (URN)978-91-8040-254-5 (ISBN)
Public defence
2022-06-10, Sal F3, Lindstedtsvägen 26, Stockholm, 14:00 (English)
Opponent
Supervisors
Note

QC 220516

Available from: 2022-05-16 Created: 2022-05-12 Last updated: 2022-06-25Bibliographically approved

Open Access in DiVA

fulltext(955 kB)114 downloads
File information
File name FULLTEXT01.pdfFile size 955 kBChecksum SHA-512
e334004071c3e761411b32f8801acbeee06d4a01beee71a0a20e0d36b975603de6ec28cb0ed814abce596dc552b3f846afc70b9f49fb484232356c671be5c136
Type fulltextMimetype application/pdf

Other links

Electronic full texthttps://proceedings.mlr.press/v151/fan22a.html

Authority records

Haasler, IsabelKarlsson, Johan

Search in DiVA

By author/editor
Haasler, IsabelKarlsson, Johan
By organisation
Mathematics (Dept.)Optimization and Systems Theory
Computational Mathematics

Search outside of DiVA

GoogleGoogle Scholar
Total: 114 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

urn-nbn

Altmetric score

urn-nbn
Total: 291 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf