A genericity property of Fr\'{e}chet sample means on Riemannian manifolds
David Groisser, Sungkyu Jung, Armin Schwartzman

TL;DR
This paper proves that for data sampled from a continuous distribution on a Riemannian manifold, the Fréchet mean almost surely avoids any measure-zero subset, establishing a strong genericity property for these means.
Contribution
The paper introduces a general geometric theorem showing Fréchet means almost surely do not lie in measure-zero subsets, extending to equivariant means on quotient manifolds.
Findings
Fréchet means almost surely avoid measure-zero sets in the manifold.
The results apply to equivariant Fréchet means on quotient manifolds.
Application demonstrated for partial scaling-rotation means of matrices.
Abstract
Let be a Riemannian manifold. If is a probability measure on given by a continuous density function, one would expect the Fr\'{e}chet means of data-samples , with respect to , to behave ``generically''; e.g. the probability that the Fr\'{e}chet mean set has any elements that lie in a given, positive-codimension submanifold, should be zero for any . Even this simplest instance of genericity does not seem to have been proven in the literature, except in special cases. The main result of this paper is a general, and stronger, genericity property: given i.i.d. absolutely continuous -valued random variables , and a subset of volume-measure zero, We also establish a companion theorem for equivariant…
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsTopological and Geometric Data Analysis
