Abstract:In this paper, we study the Radial Basis Function (RBF) approximation to differential operators on smooth tensor fields defined on closed Riemannian submanifolds of Euclidean space, identified by randomly sampled point cloud data. {The formulation in this paper leverages a fundamental fact that the covariant derivative on a submanifold is the projection of the directional derivative in the ambient Euclidean space onto the tangent space of the submanifold. To differentiate a test function (or vector field) on the submanifold with respect to the Euclidean metric, the RBF interpolation is applied to extend the function (or vector field) in the ambient Euclidean space. When the manifolds are unknown, we develop an improved second-order local SVD technique for estimating local tangent spaces on the manifold. When the classical pointwise non-symmetric RBF formulation is used to solve Laplacian eigenvalue problems, we found that while accurate estimation of the leading spectra can be obtained with large enough data, such an approximation often produces irrelevant complex-valued spectra (or pollution) as the true spectra are real-valued and positive. To avoid such an issue,} we introduce a symmetric RBF discrete approximation of the Laplacians induced by a weak formulation on appropriate Hilbert spaces. Unlike the non-symmetric approximation, this formulation guarantees non-negative real-valued spectra and the orthogonality of the eigenvectors. Theoretically, we establish the convergence of the eigenpairs of both the Laplace-Beltrami operator and Bochner Laplacian {for the symmetric formulation} in the limit of large data with convergence rates. Numerically, we provide supporting examples for approximations of the Laplace-Beltrami operator and various vector Laplacians, including the Bochner, Hodge, and Lichnerowicz Laplacians.

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is to use the Radial Basis Function (RBF) method to approximate the differential operators of smooth tensor fields defined on closed Riemannian sub - manifolds, especially the Laplace - Beltrami operator and the Bochner Laplacian. Specifically, the paper focuses on the following aspects: 1. **Estimation of differential operators**: The paper studies how to use the RBF method to approximate differential operators that act on smooth tensor fields defined on closed Riemannian sub - manifolds in Euclidean space. In particular, the paper explores how to identify these sub - manifolds through randomly sampled point - cloud data and then estimate the differential operators on this basis. 2. **Improvement of local SVD technology**: When the manifold is unknown, the paper proposes an improved second - order local Singular Value Decomposition (SVD) technique for more accurately estimating the local tangent space on the manifold. This improvement is mainly achieved by designing additional steps to correct the errors caused by curvature. 3. **Asymmetric and symmetric RBF discretizations**: The paper compares two methods, namely asymmetric RBF discretization and symmetric RBF discretization. Asymmetric RBF discretization may produce complex spectra or contaminated spectra when approximating the Laplace - Beltrami operator point - wise, while symmetric RBF discretization ensures non - negative real spectra and the orthogonality of eigenvectors. 4. **Theoretical convergence**: The paper establishes the convergence theory of symmetric RBF discretization for large amounts of data, including the convergence of eigenpairs of the Laplace - Beltrami operator and the Bochner Laplacian, and gives the convergence rate. 5. **Numerical experiments**: The paper provides numerical experiments to verify the performance of different RBF discretization methods in estimating the Laplace - Beltrami operator and various vector Laplace operators (such as Bochner, Hodge, and Lichnerowicz Laplace operators). Overall, the paper aims to evaluate the potential of the RBF method in manifold learning problems, especially its effectiveness and accuracy in dealing with high - dimensional data. Through theoretical analysis and numerical experiments, the paper shows the advantages and limitations of the RBF method in solving these problems.

Radial basis approximation of tensor fields on manifolds: From operator estimation to manifold learning

A Higher Order Local Mesh Method for Approximating Laplacians on Unknown Manifolds

Curvature corrected tangent space-based approximation of manifold-valued data

Stable and Accurate Least Squares Radial Basis Function Approximations on Bounded Domains

Collocation methods for integral fractional Laplacian and fractional PDEs based on radial basis functions

Riemannian radial distributions on Riemannian symmetric spaces: Optimal rates of convergence for parameter estimation

Convergence of Laplacian Spectra from Point Clouds

Generalized Moving Least-Squares for Solving Vector-valued PDEs on Unknown Manifolds

A Partition of Unity Method for Divergence-free or Curl-free Radial Basis Function Approximation

Polyharmonic and Related Kernels on Manifolds: Interpolation and Approximation

INTRINSIC RIEMANNIAN FUNCTIONAL DATA ANALYSIS

On the numerical rank of radial basis function kernels in high dimension

A local radial basis function method for the Laplace-Beltrami operator

Spectral stability and perturbation results for kernel differentiation matrices on the sphere

Vector Transport Free Riemannian LBFGS for Optimization on Symmetric Positive Definite Matrix Manifolds

Approximating maps into manifolds with lower curvature bounds

Second-order differential operators, stochastic differential equations and Brownian motions on embedded manifolds

A new type of radial basis functions for problems governed by partial differential equations

Radial fields on the manifolds of symmetric positive definite matrices

Learning on manifolds without manifold learning

Kernel Methods on the Riemannian Manifold of Symmetric Positive Definite Matrices