The embedding dimension of Laplacian eigenfunction maps (Q741268): Difference between revisions
From MaRDI portal
Set profile property. |
Set OpenAlex properties. |
||
Property / OpenAlex ID | |||
Property / OpenAlex ID: W3098542052 / rank | |||
Normal rank |
Revision as of 21:59, 19 March 2024
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | The embedding dimension of Laplacian eigenfunction maps |
scientific article |
Statements
The embedding dimension of Laplacian eigenfunction maps (English)
0 references
11 September 2014
0 references
For a compact Riemannian manifold \((M,g)\) with \(\{\lambda_j\}_{j \in \mathbb N}\), \(0 = \lambda_0 < \lambda_1 \leqslant \lambda_2 \leqslant \ldots \uparrow \infty\) being the spectrum of its Laplacian \(\Delta_g\) define an eigenfunction map \[ \Phi^m : M \to \mathbb R^m,\quad x \mapsto \{\varphi_j(x)\}_{1 \leqslant j \leqslant m}, \] where \(\{\varphi_j\}_{j \in \mathbb N}\) is an orthonormal basis in \(L^2(M)\) of the eigenfunctions of \(\Delta_g\), that is \(\Delta_g \varphi_j = \lambda_j \varphi_j\). The smallest number \(m\) for which \(\Phi^m\) is an embedding for every choice of the basis \(\{\varphi_j\}_{j \in \mathbb N}\) is called the maximal embedding dimension of \(M\). In the paper the author proves that for a class of closed connected \(n\)-dimensional Riemannian manifolds of unit volume, bounded from below Ricci curvature \(Ric \geqslant -(n-1)\kappa_0 g\), \(\kappa_0 \geqslant 0\), and bounded from below injectivity radius there exist a uniform maximal embedding dimension. Moreover, the same existence results if proved for two-dimensional closed surfaces with bounded Gaussian curvature \(|K| \leqslant \kappa_0\), bounded mean curvature \(|H| \leqslant H_0\), and bounded volume. Furthermore, the author describes how eigenfunction maps can be used for shape recognition problems, and explicitly compute the maximal embedding dimensions for the \(n\)-dimensional sphere and torus.
0 references
spectral embedding
0 references
eigenfunction embedding
0 references
eigenmap
0 references
diffusion map
0 references
global point signature
0 references
heat kernel embedding
0 references
shape registration
0 references
nonlinear dimensionality reduction
0 references
manifold learning
0 references