Convergence rates for empirical barycenters in metric spaces: curvature, convexity and extendible geodesics

06/07/2018
by   Adil Ahidar-Coutrix, et al.
0

This paper provides rates of convergence for empirical barycenters of a Borel probability measure on a metric space under general conditions. Our results are given in the form of sharp oracle inequalities. Our main assumption connects ideas from metric geometry to the theory of empirical processes and is studied in two meaningful scenarios. The first one is a geometrical constraint on the underlying space referred to as (k,α)-convexity, compatible with a positive upper curvature bound in the sense of Alexandrov. The second scenario considers the case of a nonnegatively curved space on which geodesics, emanating from a barycenter, can be extended. While not restricted to this setting, our results are discussed in the context of Wasserstein spaces.

READ FULL TEXT

page 1

page 2

page 3

page 4

research
06/07/2018

On the rate of convergence of empirical barycentres in metric spaces: curvature, convexity and extendible geodesics

This paper provides rates of convergence for empirical barycentres of a ...
research
08/02/2019

Fast convergence of empirical barycenters in Alexandrov spaces and the Wasserstein space

This work establishes fast rates of convergence for empirical barycenter...
research
07/18/2018

Ricci curvature for parametric statistics via optimal transport

We elaborate the notion of a Ricci curvature lower bound for parametrize...
research
04/03/2023

On the Concentration of the Minimizers of Empirical Risks

Obtaining guarantees on the convergence of the minimizers of empirical r...
research
03/02/2023

Concentration of empirical barycenters in metric spaces

Barycenters (aka Fréchet means) were introduced in statistics in the 194...
research
07/30/2022

Bump hunting through density curvature features

Bump hunting deals with finding in sample spaces meaningful data subsets...
research
11/23/2020

Level sets of depth measures and central dispersion in abstract spaces

The lens depth of a point have been recently extended to general metric ...

Please sign up or login with your details

Forgot password? Click here to reset