Abstract:Gaussian random fields are popular models for spatially varying uncertainties, arising for instance in geotechnical engineering, hydrology or image processing. A Gaussian random field is fully characterised by its mean function and covariance operator. In more complex models these can also be partially unknown. In this case we need to handle a family of Gaussian random fields indexed with hyperparameters. Sampling for a fixed configuration of hyperparameters is already very expensive due to the nonlocal nature of many classical covariance operators. Sampling from multiple configurations increases the total computational cost severely. In this report we employ parameterised Karhunen-Loève expansions for sampling. To reduce the cost we construct a reduced basis surrogate built from snapshots of Karhunen-Loève eigenvectors. In particular, we consider Matérn-type covariance operators with unknown correlation length and standard deviation. We suggest a linearisation of the covariance function and describe the associated online-offline decomposition. In numerical experiments we investigate the approximation error of the reduced eigenpairs. As an application we consider forward uncertainty propagation and Bayesian inversion with an elliptic partial differential equation where the logarithm of the diffusion coefficient is a parameterised Gaussian random field. In the Bayesian inverse problem we employ Markov chain Monte Carlo on the reduced space to generate samples from the posterior measure. All numerical experiments are conducted in 2D physical space, with non-separable covariance operators, and finite element grids with $\sim 10^4$ degrees of freedom.

What problem does this paper attempt to address?

### What problem does this paper attempt to solve? This paper aims to solve the problem of fast sampling of parameterized Gaussian Random Fields (GRFs) with high computational cost. Specifically, the author focuses on how to efficiently sample Gaussian Random Fields with unknown hyper - parameters in Bayesian inverse problems and uncertainty propagation. #### Main problem background 1. **Applications of Gaussian Random Fields**: - Gaussian Random Fields are popular models for modeling spatially varying uncertainty and are widely used in geotechnical engineering, hydrology, image processing, etc. - These random fields are fully described by a mean function and a covariance operator. When these parameters are partially unknown, a family of Gaussian Random Fields indexed by hyper - parameters needs to be dealt with. 2. **Computational challenges**: - For a fixed hyper - parameter configuration, the computational cost of sampling is already very high due to the non - locality of many classical covariance operators. - When sampling from multiple configurations is required, the total computational cost will increase significantly. - Especially in the case involving Matérn - type covariance operators, the unknown correlation length and standard deviation further increase the computational complexity. #### Methods proposed in the paper To reduce the computational cost, the author proposes the following methods: 1. **Parameterized Karhunen - Loève expansion**: - Use the parameterized Karhunen - Loève (KL) expansion to sample Gaussian Random Fields. - The KL expansion represents the random field as a linear combination of a set of orthogonal basis functions, and its coefficients are independent standard normal distribution random variables. 2. **Reduced Basis Method**: - Construct a reduced - basis surrogate model based on KL eigenvector snapshots to reduce the computational cost. - Specifically, for different hyper - parameter values, pre - compute some KL eigenpairs and use the linear combination of these eigenpairs to approximate new eigenpairs. - This method can significantly reduce the amount of computation required to solve the eigenvalue problem because it only needs to be solved in a smaller space. 3. **Online - offline decomposition**: - Propose a linearization method of the covariance function and describe the related online - offline decomposition. - The offline phase computes and stores the reduced - basis space, while the online phase uses these pre - computed results to quickly generate samples. #### Application examples - **Forward uncertainty propagation**: Study the influence of uncertain model inputs on model outputs. - **Bayesian inverse problems**: Use observed data to update the prior distribution of random inputs to obtain the posterior distribution. - In Bayesian inverse problems, the Markov chain Monte Carlo (MCMC) method is used to generate samples of the posterior measure in the reduced - basis space. Through these methods, the paper aims to provide an efficient sampling technique that can significantly reduce the computational cost while maintaining accuracy, so as to better handle the problem of Gaussian Random Fields with unknown hyper - parameters.

Fast sampling of parameterised Gaussian random fields

A Surrogate Modelling Approach Based on Nonlinear Dimension Reduction for Uncertainty Quantification in Groundwater Flow Models

Residual Gaussian Process: A Tractable Nonparametric Bayesian Emulator for Multi-Fidelity Simulations

A Multilevel, Hierarchical Sampling Technique for Spatially Correlated Random Fields

A Fast GP Regression Method Using Banded Sparsification of Inverse Covariance

Multilevel approximation of Gaussian random fields: Covariance compression, estimation, and spatial prediction

An explicit link between Gaussian fields and Gaussian Markov random fields: the stochastic partial differential equation approach

Fast Adaptive Fourier Integration for Spectral Densities of Gaussian Processes

A Chebyshev Polynomial-Based Galerkin Method for the Discretization of Spatially Varying Random Properties

A Galerkin Isogeometric Method for Karhunen-Loeve Approximation of Random Fields

Sampling Gaussian stationary random fields: A stochastic realization approach

Bayesian Learning of Orthogonal Embeddings for Multi-Fidelity Gaussian Processes

Non-stationary Gaussian random fields on hypersurfaces: Sampling and strong error analysis

On the Numerical Approximation of the Karhunen-Loève Expansion for Random Fields with Random Discrete Data

Solving High Frequency and Multi-Scale PDEs with Gaussian Processes

Uniform approximation of common Gaussian process kernels using equispaced Fourier grids

Numerical Simulation of Random Fields with a High-Order Polynomial Based Ritz–Galerkin Approach

Simulation of Non-Stationary Non-Gaussian Random Fields from Sparse Measurements Using Bayesian Compressive Sampling and Karhunen-Loève Expansion

Dirichlet-Neumann Averaging: The DNA of Efficient Gaussian Process Simulation

Modified Linear Estimation Method for Generating Multi-Dimensional Multi-Variate Gaussian Field in Modelling Material Properties