Analysis and Operator Theory Seminar - Mikhail Ostrovskii

Tue, March 30, 2027
1:50 pm - 2:50 pm
TBD

Mikhail Ostrovskii
St. John's University

Title
Finite-Dimensional Transportation Cost Spaces and Their Applications

Abstract
Given a finite metric space $X$, consider the vector space of signed real measures $\mu$ on $X$ with $\mu(X)=0$. The minimum cost of transporting the positive part of $\mu$ to its negative part, with cost per unit mass equal to the distance traveled, defines the \emph{transportation cost norm}. We denote the resulting normed space by $\TC(X)$. These spaces are also known as Arens–Eells spaces or Lipschitz-free spaces; their norm is the Kantorovich–Rubinstein norm, and the corresponding distance between probability measures is the earth mover’s distance, or $1$-Wasserstein distance.
A central problem, motivated in part by applications to efficient image comparison in computer vision, is to determine how well $\TC(X)$ embeds into $L_1$. For a metric space $(W,d)$, its \emph{$L_1$-distortion} $c_1(W)$ is the infimum of all $C\geq 1$ for which there exists a map $F\colon W\to L_1(0,1)$ satisfying
[
d(u,v)\leq |F(u)-F(v)|_1\leq C\,d(u,v)
\qquad (u,v\in W).
]
After introducing the basic properties of transportation cost spaces, I will discuss recent results on $c_1(\TC(X))$. The main results give lower bounds of optimal asymptotic order when $X$ is a square planar grid, a high-dimensional grid, or a graph from a recursive family, including diamond graphs, Laakso graphs, and their generalizations.


The results for square grids and recursive graph families are joint work with Chris Gartland. The results for high-dimensional grids are joint work with Chris Gartland, Yuval Rabani, and Robert Young.

For More Information About the Seminar