A direct approach for function approximation on data defined manifolds
From MaRDI portal
Publication:2057766
Abstract: In much of the literature on function approximation by deep networks, the function is assumed to be defined on some known domain, such as a cube or a sphere. In practice, the data might not be dense on these domains, and therefore, the approximation theory results are observed to be too conservative. In manifold learning, one assumes instead that the data is sampled from an unknown manifold; i.e., the manifold is defined by the data itself. Function approximation on this unknown manifold is then a two stage procedure: first, one approximates the Laplace-Beltrami operator (and its eigen-decomposition) on this manifold using a graph Laplacian, and next, approximates the target function using the eigen-functions. Alternatively, one estimates first some atlas on the manifold and then uses local approximation techniques based on the local coordinate charts. In this paper, we propose a more direct approach to function approximation on emph{unknown}, data defined manifolds without computing the eigen-decomposition of some operator or an atlas for the manifold, and without any kind of training in the classical sense. Our constructions are universal; i.e., do not require the knowledge of any prior on the target function other than continuity on the manifold. We estimate the degree of approximation. For smooth functions, the estimates do not suffer from the so-called saturation phenomenon. We demonstrate via a property called good propagation of errors how the results can be lifted for function approximation using deep networks where each channel evaluates a Gaussian network on a possibly unknown manifold.
Recommendations
- Provable approximation properties for deep neural networks
- Approximation of functions over manifolds: a moving least-squares approach
- Function approximation by deep networks
- Estimation of a regression function on a manifold by fully connected deep neural networks
- Eignets for function approximation on manifolds
Cites work
- A Fourier-invariant method for locating point-masses and computing their attributes
- A generalized diffusion frame for parsimonious representation of functions on data defined manifolds
- A unified framework for harmonic analysis of functions on directed graphs and changing data
- A unified method for super-resolution recovery and real exponential-sum separation
- An analysis of training and generalization errors in shallow and deep networks
- Bigeometric organization of deep nets
- Concentration inequalities. A nonasymptotic theory of independence
- Deep vs. shallow networks: an approximation theory perspective
- Diffusion polynomial frames on metric measure spaces
- Eignets for function approximation on manifolds
- From graph to manifold Laplacian: the convergence rate
- scientific article; zbMATH DE number 1190534 (Why is no real title available?)
- scientific article; zbMATH DE number 52737 (Why is no real title available?)
- scientific article; zbMATH DE number 3477793 (Why is no real title available?)
- scientific article; zbMATH DE number 1231230 (Why is no real title available?)
- scientific article; zbMATH DE number 477682 (Why is no real title available?)
- scientific article; zbMATH DE number 1972818 (Why is no real title available?)
- scientific article; zbMATH DE number 823880 (Why is no real title available?)
- scientific article; zbMATH DE number 2204742 (Why is no real title available?)
- scientific article; zbMATH DE number 3108589 (Why is no real title available?)
- Laplacian Eigenmaps for Dimensionality Reduction and Data Representation
- Learning Theory
- Local approximation using Hermite functions
- Marcinkiewicz-Zygmund measures on manifolds
- Mean Convergence of Expansions in Laguerre and Hermite Series
- Networks and the best approximation property
- On the mathematical foundations of learning
- Semi-supervised learning on Riemannian manifolds
- Towards a theoretical foundation for Laplacian-based manifold methods
- Universal local parametrizations via heat kernels and eigenfunctions of the Laplacian
- When is approximation by Gaussian networks necessarily a linear process?
Cited in
(6)- Construct Deep Neural Networks based on Direct Sampling Methods for Solving Electrical Impedance Tomography
- Tikhonov regularization for polynomial approximation problems in Gauss quadrature points
- A generalized diffusion frame for parsimonious representation of functions on data defined manifolds
- A deep network construction that adapts to intrinsic dimensionality beyond the domain
- Parameter choice strategies for regularized least squares approximation of noisy continuous functions on the unit circle
- Eignets for function approximation on manifolds
This page was built for publication: A direct approach for function approximation on data defined manifolds
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2057766)
