KRIGING IN TENSOR TRAIN DATA FORMAT

@article{Dolgov2019KRIGINGIT,
  title={KRIGING IN TENSOR TRAIN DATA FORMAT},
  author={Sergey V. Dolgov and Alexander Litvinenko and Dishi Liu},
  journal={Proceedings of the 3rd International Conference on Uncertainty Quantification in Computational Sciences and Engineering (UNCECOMP 2019)},
  year={2019}
}
  • S. Dolgov, A. Litvinenko, Dishi Liu
  • Published 21 April 2019
  • Computer Science
  • Proceedings of the 3rd International Conference on Uncertainty Quantification in Computational Sciences and Engineering (UNCECOMP 2019)
Combination of low-tensor rank techniques and the Fast Fourier transform (FFT) based methods had turned out to be prominent in accelerating various statistical operations such as Kriging, computing conditional covariance, geostatistical optimal design, and others. However, the approximation of a full tensor by its low-rank format can be computationally formidable. In this work, we incorporate the robust Tensor Train (TT) approximation of covariance matrices and the efficient TT-Cross algorithm… 

Figures and Tables from this paper

Computing f-Divergences and Distances of High-Dimensional Probability Density Functions - Low-Rank Tensor Approximations
TLDR
The connection between the low-rank approximability of the d-dimensional random variable, its pcf , and its pdf is investigated, and it is shown how to go from the pcf or functional representation to the pdf, to reduce the computational complexity and storage cost from O(nd) to O(dnrα).

References

SHOWING 1-10 OF 62 REFERENCES
Kriging and Spatial Design Accelerated by Orders of Magnitude: Combining Low-Rank Covariance Approximations with FFT-Techniques
Computational power poses heavy limitations to the achievable problem size for Kriging. In separate research lines, Kriging algorithms based on FFT, the separability of certain covariance functions,
Tucker Tensor Analysis of Matérn Functions in Spatial Statistics
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
The Alternating Linear Scheme for Tensor Optimization in the Tensor Train Format
TLDR
This article shows how optimization tasks can be treated in the TT format by a generalization of the well-known alternating least squares (ALS) algorithm and by a modified approach (MALS) that enables dynamical rank adaptation.
Multilevel Toeplitz Matrices Generated by Tensor-Structured Vectors and Convolution with Logarithmic Complexity
TLDR
It is shown how the multilevel Toeplitz matrix generated by $\boldsymbol{x}$ can be obtained in the QTT format with ranks bounded by $2p$ in $\mathcal{O}\left(dp^{2} \log n\right)$ operations.
Application of FFT-based Algorithms for Large-Scale Universal Kriging Problems
Looking at kriging problems with huge numbers of estimation points and measurements, computational power and storage capacities often pose heavy limitations to the maximum manageable problem size. In
TT-cross approximation for multidimensional arrays
Likelihood approximation with hierarchical matrices for large spatial datasets
Efficient methods for large‐scale linear inversion using a geostatistical approach
TLDR
Hierarchical Matrices is used to reduce the complexity of forming approximate matrix‐vector products involving the Covariance matrices in log linear complexity for an arbitrary distribution of points and a wide variety of generalized covariance functions and solves iteratively using a matrix‐free Krylov subspace approach.
Efficient Computation of Linearized Cross-Covariance and Auto-Covariance Matrices of Interdependent Quantities
TLDR
This paper summarizes and extends a collection of highly efficient spectral methods to compute these matrices, based on circulant embedding and the fast Fourier transform, applicable whenever the unknowns are a stationary random variable discretized on a regular equispaced grid.
FFT-based homogenisation accelerated by low-rank tensor approximations
...
...