Tucker tensor analysis of Matérn functions in spatial statistics

Alexander Litvinenko, David Keyes, Venera Khoromskaia, Boris N. Khoromskij, and Hermann G. Matthies

Contact the author: Please use for correspondence this email.
Submission date: 08. Mar. 2018
Pages: 24
published in: Computational methods in applied mathematics, 19 (2019) 1, p. 101-122 
DOI number (of the published article): 10.1515/cmam-2018-0022
Bibtex
MSC-Numbers: 60H15, 60H35, 65N25
Keywords and phrases: low-rank tensor approximationform, geostatistical optimal design, Kriging, Mat\'{e}rn covariance, Hilbert tensor, Kalman filter, Bayesian update, loglikelihood surrogate
Download full preprint: PDF (871 kB)

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 Matérn- 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 O(nˆd) to a linear scaling O(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, ∥x − y∥.