Parameter estimation in high dimensional Gaussian distributions

Authors:
Erlend Aune;Daniel P. Simpson;Jo Eidsvik
Affiliations:
Norwegian University of Science and Technology, Trondheim, Norway;Norwegian University of Science and Technology, Trondheim, Norway;Norwegian University of Science and Technology, Trondheim, Norway
Venue:
Statistics and Computing
Year:
2014

Citing 13
Cited 0

Some large-scale matrix computation problems

Journal of Computational and Applied Mathematics - Special issue on TICAM symposium
Parallel Preconditioning with Sparse Approximate Inverses

SIAM Journal on Scientific Computing
A Fast and High Quality Multilevel Scheme for Partitioning Irregular Graphs

SIAM Journal on Scientific Computing
Modifying a Sparse Cholesky Factorization

SIAM Journal on Matrix Analysis and Applications
Iterative Methods for Sparse Linear Systems

Iterative Methods for Sparse Linear Systems
BiCGStab(l) for families of shifted linear systems

Computing
Gaussian Markov Random Fields: Theory And Applications (Monographs on Statistics and Applied Probability)

Gaussian Markov Random Fields: Theory And Applications (Monographs on Statistics and Applied Probability)
An estimator for the diagonal of a matrix

Applied Numerical Mathematics
Functions of Matrices: Theory and Computation (Other Titles in Applied Mathematics)

Functions of Matrices: Theory and Computation (Other Titles in Applied Mathematics)
Algorithm 887: CHOLMOD, Supernodal Sparse Cholesky Factorization and Update/Downdate

ACM Transactions on Mathematical Software (TOMS)
Computing $A^\alpha, \log(A)$, and Related Matrix Functions by Contour Integrals

SIAM Journal on Numerical Analysis
A Matrix-free Approach for Solving the Parametric Gaussian Process Maximum Likelihood Problem

SIAM Journal on Scientific Computing
Iterative numerical methods for sampling from high dimensional Gaussian distributions

Statistics and Computing

Quantified Score

Hi-index	0.00

Visualization

Abstract

In order to compute the log-likelihood for high dimensional Gaussian models, it is necessary to compute the determinant of the large, sparse, symmetric positive definite precision matrix. Traditional methods for evaluating the log-likelihood, which are typically based on Cholesky factorisations, are not feasible for very large models due to the massive memory requirements. We present a novel approach for evaluating such likelihoods that only requires the computation of matrix-vector products. In this approach we utilise matrix functions, Krylov subspaces, and probing vectors to construct an iterative numerical method for computing the log-likelihood.