Tucker tensor analysis of Matérn functions in spatial statistics

MaRDI QIDQ2324355zbMATH OpenFDO

Authors Alexander Litvinenko, D. E. Keyes, V. Khoromskaia, Boris N. Khoromskij, Hermann G. Matthies

Publication date 11 September 2019

Published in Computational Methods in Applied Mathematics (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1711.06874

zbMATH Keywords

Bayesian updateFourier transformgeostatistical optimal designHilbert tensorKalman filterKrigingloglikelihood surrogatelow-rank tensor approximationMatérn covariance

Mathematics Subject Classification ID

Stochastic partial differential equations (aspects of stochastic analysis) (60H15) Computational methods for stochastic equations (aspects of stochastic analysis) (60H35) Numerical methods for eigenvalue problems for boundary value problems involving PDEs (65N25)

Abstract: In this work, we describe advanced numerical tools for working with multivariate functions and for the analysis of large data sets. These tools will drastically reduce the required computing time and the storage cost, and, therefore, will allow us to consider much larger data sets or finer meshes. Covariance matrices are crucial in spatio-temporal statistical tasks, but are often very expensive to compute and store, especially in 3D. Therefore, we approximate covariance functions by cheap surrogates in a low-rank tensor format. We apply the Tucker and canonical tensor decompositions to a family of Matern- and Slater-type functions with varying parameters and demonstrate numerically that their approximations exhibit exponentially fast convergence. We prove the exponential convergence of the Tucker and canonical approximations in tensor rank parameters. Several statistical operations are performed in this low-rank tensor format, including evaluating the conditional covariance matrix, spatially averaged estimation variance, computing a quadratic form, determinant, trace, loglikelihood, inverse, and Cholesky decomposition of a large covariance matrix. Low-rank tensor approximations reduce the computing and storage costs essentially. For example, the storage cost is reduced from an exponential $mathcalO(n^d)$ to a linear scaling $mathcalO(drn)$, where $d$ is the spatial dimension, $n$ is the number of mesh points in one direction, and $r$ is the tensor rank. Prerequisites for applicability of the proposed techniques are the assumptions that the data, locations, and measurements lie on a tensor (axes-parallel) grid and that the covariance function depends on a distance, $Vertx-yVert$.

Recommendations

• Tensor spaces and numerical tensor calculus
• Efficient randomized tensor-based algorithms for function approximation and low-rank kernel interactions
• Linear-Cost Covariance Functions for Gaussian Random Fields
• Approximation of Spatio-Temporal Random Processes Using Tensor Decomposition
• Numerical tensor calculus

Cites work

• A literature survey of low-rank tensor approximation techniques
• A Multilinear Singular Value Decomposition
• A sparse \({\mathcal H}\)-matrix arithmetic. II: Application to multi-dimensional problems
• A sparse matrix arithmetic based on \({\mathfrak H}\)-matrices. I: Introduction to \({\mathfrak H}\)-matrices
• A unifying view of sparse approximate Gaussian process regression
• Aggregation-cokriging for highly multivariate spatial data
• Application of hierarchical matrices for computing the Karhunen-Loève expansion
• Approximating Likelihoods for Large Spatial Data Sets
• Collecting Spatial Data
• Computation of the Hartree-Fock exchange by the tensor-structured methods
• Data-sparse approximation to a class of operator-valued functions
• Difference filter preconditioning for large covariance matrices
• Efficient approximation of random fields for numerical applications.
• Fast spatial Gaussian process maximum likelihood estimation via skeletonization factorizations
• Fast tensor method for summation of long-range potentials on 3D lattices with defects.
• Gaussian processes for machine learning.
• Geometrical structures of FIR manifold and multichannel blind deconvolution
• Hierarchical matrices: algorithms and analysis
• Hierarchical tensor-product approximation to the inverse and related operators for high-dimensional elliptic problems
• scientific article; zbMATH DE number 435359 (Why is no real title available?)
• scientific article; zbMATH DE number 4071579 (Why is no real title available?)
• scientific article; zbMATH DE number 1313649 (Why is no real title available?)
• Kriging and spatial design accelerated by orders of magnitude: combining low-rank covariance approximations with FFT-techniques
• Large-scale stochastic linear inversion using hierarchical matrices. Illustrated with an application to crosswell tomography in seismic imaging
• Likelihood approximation with hierarchical matrices for large spatial datasets
• Low rank Tucker-type tensor approximation to classical potentials
• Low-rank Kronecker-product approximation to multi-dimensional nonlocal operators I. Separable approximation of multi-variate functions
• Low-rank Kronecker-product approximation to multi-dimensional nonlocal operators II. HKT representation of certain operators
• Low-rank quadrature-based tensor approximation of the Galerkin projected Newton/Yukawa kernels
• Matheron's theory of regionalised variables. Edited by Vera Pawlowsky-Glahn and Jean Serra. Translated from the French by Charles Huijbregts
• Matrices with hierarchical low-rank structures
• Measures of parameter uncertainty in geostatistical estimation and geostatistical optimal design
• Modeling and prediction of multivariate space-time random fields
• Multigrid accelerated tensor approximation of function related multidimensional arrays
• Orthogonal tensor decompositions
• Polynomial Chaos Expansion of Random Coefficients and the Solution of Stochastic Partial Differential Equations in the Tensor Train Format
• Spatial sampling design and covariance-robust minimax prediction based on convex design ideas
• Spatial variation. 2nd ed
• Structured Rank-(r1, . . . , rd) Decomposition of Function-related Tensors in R_D
• Superfast Fourier transform using QTT approximation
• Tensor Decompositions and Applications
• Tensor numerical methods for multidimensional PDEs: theoretical analysis and initial applications
• Tensor rank is NP-complete
• Tensor spaces and numerical tensor calculus
• Tensor-train decomposition
• The density-matrix renormalization group in the age of matrix product states

Cited in

(6)

• Prospects of tensor-based numerical modeling of the collective electrostatics in many-particle systems
• Efficient randomized tensor-based algorithms for function approximation and low-rank kernel interactions
• Range-separated tensor decomposition of the discretized Dirac delta and elliptic operator inverse
• Editorial. Tensor numerical methods: actual theory and recent applications
• Stochastic modelling of symmetric positive definite material tensors
• Sum of Kronecker products representation and its Cholesky factorization for spatial covariance matrices from large grids

Describes a project that uses

Uses Software

• HLIBCov
• ICALAB

This page was built for publication: Tucker tensor analysis of Matérn functions in spatial statistics

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2324355)