About

The last few years have seen the rapid development of mathematical methods for modeling structured data coming from biology, chemistry, network science, natural language processing, and computer vision applications. Recently developed tools and cutting-edge methodologies coming from the theory of optimal transport have proved to be particularly successful for these tasks. A striking feature of much of this recent work is the application of new theoretical and computational techniques for comparing probability distributions defined on spaces with complex structure, such as graphs, Riemannian manifolds and more general metric spaces.

This workshop aims to bring together leading computer scientists, mathematicians, AI researchers and practitioners from theoretical and applied communities to share ideas, promote advanced work, and foster collaboration. It will provide a premier interdisciplinary forum to discuss the most recent trends, innovations, applications, and challenges of optimal transport and structured data modeling.

We invite talks and submissions focusing on computational and theoretical aspects of optimal transport and applications of structured data modeling, especially on optimal transport-based machine learning methods and on optimal transport applied to distributions on incomparable spaces. More topics are listed in the call for papers.

Schedule (Pacific Time)

America-Europe Session (8 am - 12 pm, Feb 28, PST)

8:00 am - 8:10 am: Opening

8:10 am - 9:00 am: Long Talk, Optimal Transport in Single-Cell Biology: Challenges and Opportunities

Caroline Uhler, MIT

9:00 am - 9:50 am: Long Talk, Scaling Optimal Transport for High Dimensional Learning

Gabriel Peyré, CNRS and Ecole Normale Supérieure

9:50 am - 10:10 am: Short Break

10:10 am - 10:35 am: Short Talk, The (Fused) Gromov-Wasserstein Framework as a Tool for Learning on Structured Data

Titouan Vayer, ENS Lyon

10:35 am - 10:50 am: Spotlight, Functional Optimal Transport: map estimation and domain adaptation for functional data

10:50 am - 11:05 am: Spotlight, Ocean Mover's Distance: Using Optimal Transport for Analyzing Oceanographic Data

11:05 am - 11:20 am: Spotlight, The Gene Mover's Distance: Single-Cell Similarity via Optimal Transport

11:20 pm - 11:35 pm: Spotlight, Center-Outward Sign- and Rank-Based Quadrant, Spearman, and Kendall Tests for Multivariate Independence

America-Asia Session (4 pm - 8 pm, Feb 28, PST)

4:00 pm - 4:10 pm: Opening

4:10 pm - 5:00 pm: Long Talk, Weisfeiler-Lehman meets Gromov-Wasserstein

Yusu Wang, UCSD

5:00 pm - 5:25 pm: Short Talk, On Scalability of Optimal Transport with Tree/Graph Metric

Tam Le, RIKEN

5:25 pm - 5:50 pm: Short Talk, Differentiable Hierarchical Optimal Transport for Robust Multi-view Learning

Dixin Luo, Beijing Institute of Technology

5:50 pm - 6:10 pm: Short Break

6:10 pm - 6:25 pm: Spotlight, Combining Reinforcement Learning and Optimal Transport for the Traveling Salesman Problem

6:25 pm - 6:40 pm: Spotlight, Exploiting Problem Structure in Deep Declarative Networks: Two Case Studies

6:40 pm - 6:55 pm: Spotlight, Obtaining Dyadic Fairness by Optimal Transport

6:55 pm - 7:00 pm Closing

Long Talks (40 mins + 10 mins QA)

Scaling Optimal Transport for High Dimensional Learning

Gabriel Peyré, (Mathematics, CNRS Senior Researcher)

Weisfeiler-Lehman meets Gromov-Wasserstein

Yusu Wang, (Mathematics, Professor in CSE, UCSD)

Optimal Transport in Single-Cell Biology: Challenges and Opportunities

Caroline Uhler, (Statistics and CS, Associate Professor in EECS and IDSS, MIT)

Short Talks (20 mins + 5 mins QA)

The (Fused) Gromov-Wasserstein Framework as a Tool for Learning on Structured Data

Titouan Vayer, (Mathematics, Postdoctoral Researcher at ENS Lyon)

On Scalability of Optimal Transport with Tree/Graph Metric

Tam Le, (Computer Science, Research Scientist at RIKEN)

Differentiable Hierarchical Optimal Transport for Robust Multi-view Learning

Dixin Luo, (Computer Science, Assistant Professor in CS, Beijing Institute of Technology)

Accepted Papers (with Spotlight Presentations, 12 mins + 3 mins QA)

Obtaining Dyadic Fairness by Optimal Transport [paper]

Moyi Yang (East Chine Normal University); Junjie Sheng (East China Normal University); Xiangfeng Wang (East China Normal University); Wenyan Liu (East China Normal University); Bo Jin (East China Normal University/Shanghai Research Institute for Intelligent Autonomous Systems); Jun Wang (East China Normal University); Hongyuan Zha (Chinese University of Hong Kong, Shenzhen)

Combining Reinforcement Learning and Optimal Transport for the Traveling Salesman Problem [paper]

Yong Liang Goh (National University of Singapore); Wee Sun Lee (National University of Singapore); Xavier Bresson (National University of Singapore); Thomas Laurent (Loyola Marymount University); Xiang Hui Nicholas Lim (NUS-Grab)

Functional Optimal Transport: map estimation and domain adaptation for functional data [paper]

Jiacheng Zhu (Carnegie Mellon University); Aritra Guha (Duke University); Dat Do (University of Michigan); Mengdi Xu (Carnegie Mellon University); XuanLong Nguyen (University of Michigan); Ding Zhao (Carnegie Mellon University)

Ocean Mover's Distance: Using Optimal Transport for Analyzing Oceanographic Data [paper]

Sangwon Hyun (University of Southern California); Aditya Mishra (Flatiron Institute); Christopher Follett (Massachusetts Institute of Technology); Bror Jonsson (Plymouth Marine Laboratory); Gemma Kulk (Plymouth Marine Laboratory); Gael Forget (Massachusetts Institute of Technology); Marie-Fanny Racault (Plymouth Marine Laboratory); Thoams Jackson (Plymouth Marine Laboratory); Stephanie Dutkiewicz (Massachusetts Institute of Technology); Christian L. Müller (Center for Computational Mathematics, Flatiron Institute); Jacob Bien (University of Southern California)

Exploiting Problem Structure in Deep Declarative Networks: Two Case Studies [paper] [slides]

Stephen Gould (Australian National University, Australia); Dylan Campbell (University of Oxford); Yizhak Ben-Shabat (Technion); Chamin P Hewa Koneputugodage (Australian National University); Zhiwei Xu (Australian National University)

Center-Outward Sign- and Rank-Based Quadrant, Spearman, and Kendall Tests for Multivariate Independence [paper]

Hongjian Shi (Technical University of Munich); Mathias Drton (Technical University of Munich); Marc Hallin (Université Libre de Bruxelles); Fang Han (University of Washington)

The Gene Mover's Distance: Single-Cell Similarity via Optimal Transport [paper]

Stefano Gualandi (University of Pavia); Eleonora Vercesi (University of Pavia); Andrea Codegoni (Università degli Studi di Pavia); Gianni Guidetti (University of Pavia); Giovanna Nicora (University of Pavia); Riccardo Bellazzi (University of Pavia)

Call for Papers

We welcome contributions of both technical and perspective papers from a wide range of topics, including but not limited to the following topics of interest:

1. Theoretical and Computational Optimal Transport
   • Optimal transport theory, including statistical and geometric aspects
   • Gromov-Wasserstein distance and its variants
   • Theory of unbalanced or partial optimal transport
   • Computational methods of optimal transport
2. Optimal Transport-Driven Machine Learning
   • Bayesian inference for/with optimal transport
   • Gromovization of machine learning methods
   • Optimal transport-driven generative modeling
   • Optimal transport-based machine learning, such as metric, relational, multi-modality and multi-task learning
   • Trustworthy machine learning from the perspective of optimal transport
3. Optimal Transport-Based Structured Data Modeling
   • Optimal transport-based analysis of structured data, such as networks, meshes, topological data, sequential data or manifold-valued data
   • Specific optimal transport-based applications such as graph analysis, natural language processing, computer vision, bioinformatics, or analysis of molecular data
   • Theory of global geometry of spaces of structured data, such as metric measure spaces, kernel functions, manifolds or networks

PAPER SUBMISSION GUIDLINES

We invite the submission of papers with 4-6 pages. References will not count towards the page limit. Papers must be in PDF format, in English, and formatted according to the AAAI template.  Submissions will be peer-reviewed, single-blinded, and assessed based on their novelty, technical quality, significance, clarity, and relevance regarding the workshop topics. Submissions introducing interesting experimental phenomena and open problems of optimal transport and structured data modeling are welcome as well.  All accepted papers will be posted on the workshop website.

Submission is permitted for papers that are under review, posted in arXiv, or planned to be submitted elsewhere. To encourage original works, submissions that are already accepted for another conference or a journal are not accepted. This policy also applies to submissions that overlap substantially in technical content with papers previously published or accepted.

The submission website is https://cmt3.research.microsoft.com/OTSDM2022.

Important Days

All time are 23:59, AoE (Anywhere on Earth)
November 26, 2021: Submission due
January 7, 2022: Paper notification
February 28, 2022: Workshop day
March 1, 2022: Camera-ready due