Back to search

Cornell University

The Gromov-Wasserstein Alignment and Optimal Transport: Geometry over Probability Distributions and Applications

Abstract

dc:description.abstract

At the intersection of several pure and applied mathematical disciplines, optimal transport (OT) has been drawing surprising connections across geometry, analysis and statistics, while also offering powerful computational methods in data science and machine learning. The increasing interest in recent years in heterogeneous data has, however, not been quite compatible with OT, and requires systematic treatment of a closely related but significantly harder object called the Gromov-Wasserstein (GW) alignment. The GW framework inherently only utilizes the intrinsic structure of data, modeled as metric measure spaces, and has seen widespread applications. Despite its compelling mathematical formulation, GW encompasses several stark structural differences, prohibiting the extension of much of classical OT studies to the GW problem, while hindering the development of better computational methods for GW. This thesis addresses these gaps, motivated by establishing the transport aspect of GW. The first contribution is a novel duality theory for GW and the first statistical estimation rate of the (2,2)-GW distance and its entropic regularized variant, with matching lower bounds. For better applicability, the second contribution of this thesis seeks statistically and computationally efficient heterogeneous methods under GW's framework. The third contribution delves deeper into drawing parallels from OT's Riemannian structure to GW, leading to novel gradient flow and dynamical formulation.

Degree

thesis:*
Name thesis:degree_name
Ph. D., Applied Mathematics
Level thesis:degree_level
Doctor of Philosophy
Discipline thesis:degree_discipline
Applied Mathematics
Grantor
Cornell University
Year dc:date.issued
2025

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Zhang, Zhengxin
Chair dc:contributor.chair
  • Goldfeld, Ziv
Committee members dc:contributor.committeemember
  • Kato, Kengo
  • Saloff-Coste, Laurent

Subjects

dc:subject × 3

Rights

Language dc:language.iso
en

Identifiers

dc:identifier.*
Dc Identifier Bibid
17423387
Dc Identifier Other
ProQuest Submission ID: 15372
ProQuest Publication ID: 32287058

Chain of custody

source
Harvested from
Cornell University
Base URL
ecommons.cornell.edu/server/oai/request
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
citation

Zhang, Zhengxin. The Gromov-Wasserstein Alignment and Optimal Transport: Geometry over Probability Distributions and Applications. Doctor of Philosophy thesis, Cornell University, 2025. https://hdl.handle.net/1813/121021