English

Minimax Boundary Estimation and Estimation with Boundary

Statistics Theory 2023-03-13 v2 Statistics Theory

Abstract

We derive non-asymptotic minimax bounds for the Hausdorff estimation of dd-dimensional submanifolds MRDM \subset \mathbb{R}^D with (possibly) non-empty boundary M\partial M. The model reunites and extends the most prevalent C2\mathcal{C}^2-type set estimation models: manifolds without boundary, and full-dimensional domains. We consider both the estimation of the manifold MM itself and that of its boundary M\partial M if non-empty. Given nn samples, the minimax rates are of order O((logn/n)2/d)O\bigl((\log n/n)^{2/d}\bigr) if M=\partial M = \emptyset and O((logn/n)2/(d+1))O\bigl((\log n/n)^{2/(d+1)}\bigr) if M\partial M \neq \emptyset, up to logarithmic factors. In the process, we develop a Voronoi-based procedure that allows to identify enough points O((logn/n)2/(d+1))O\bigl((\log n/n)^{2/(d+1)}\bigr)-close to M\partial M for reconstructing it.

Keywords

Cite

@article{arxiv.2108.03135,
  title  = {Minimax Boundary Estimation and Estimation with Boundary},
  author = {Eddie Aamari and Catherine Aaron and Clément Levrard},
  journal= {arXiv preprint arXiv:2108.03135},
  year   = {2023}
}
R2 v1 2026-06-24T04:53:36.680Z