Nonparametric Analysis of Non-Euclidean Data on Shapes and Images

Abstract

The article presents some of the basic theory for nonparametric inference on non-Euclidean spaces using Frechet means that has been developed during the past two decades. Included are recent results on the asymptotic distribution theory of sample Frechet means on such spaces, especially differentiable and Riemannian manifolds. Apart from this main theme and its applications, a nonparametric Bayes theory on Riemannian manifolds is outlined for the purpose of density estimation and classification. A final section briefly discusses the problem of machine vision, or robotic recognition of images as Riemannian manifolds.