Fast spatial Gaussian process maximum likelihood estimation via skeletonization factorizations
From MaRDI portal
Abstract: Maximum likelihood estimation for parameter-fitting given observations from a Gaussian process in space is a computationally-demanding task that restricts the use of such methods to moderately-sized datasets. We present a framework for unstructured observations in two spatial dimensions that allows for evaluation of the log-likelihood and its gradient (i.e., the score equations) in time under certain assumptions, where is the number of observations. Our method relies on the skeletonization procedure described by Martinsson & Rokhlin in the form of the recursive skeletonization factorization of Ho & Ying. Combining this with an adaptation of the matrix peeling algorithm of Lin et al. for constructing -matrix representations of black-box operators, we obtain a framework that can be used in the context of any first-order optimization routine to quickly and accurately compute maximum-likelihood estimates.
Recommendations
- A matrix-free approach for solving the parametric Gaussian process maximum likelihood problem
- Gaussian process learning via Fisher scoring of Vecchia's approximation
- Local inversion-free estimation of spatial Gaussian processes
- Stochastic approximation of score functions for Gaussian processes
- Scalable computations for nonstationary Gaussian processes
Cites work
- A Fast $ULV$ Decomposition Solver for Hierarchically Semiseparable Representations
- A Fast Solver for HSS Representations via Sparse Matrices
- A Full Scale Approximation of Covariance Functions for Large Spatial Data Sets
- A direct solver with O(N) complexity for integral equations on one-dimensional domains
- A fast direct solver for boundary integral equations in two dimensions
- A fast direct solver for structured linear systems by recursive skeletonization
- A fast randomized algorithm for overdetermined linear least-squares regression
- A matrix-free approach for solving the parametric Gaussian process maximum likelihood problem
- A recursive skeletonization factorization based on strong admissibility
- 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 stochastic estimator of the trace of the influence matrix for laplacian smoothing splines
- An \(O(N)\) direct solver for integral equations on the plane
- An explicit link between Gaussian fields and Gaussian Markov random fields: the stochastic partial differential equation approach
- Application of hierarchical matrices for computing the Karhunen-Loève expansion
- Approximating Likelihoods for Large Spatial Data Sets
- Data-sparse approximation by adaptive \({\mathcal H}^2\)-matrices
- Difference filter preconditioning for large covariance matrices
- Fast algorithms for hierarchically semiseparable matrices
- Fast construction of hierarchical matrix representation from matrix-vector multiplication
- Fast direct solvers for integral equations in complex three-dimensional domains
- Finding structure with randomness: probabilistic algorithms for constructing approximate matrix decompositions
- Fixed Rank Kriging for Very Large Spatial Data Sets
- Hierarchical interpolative factorization for elliptic operators: differential equations
- Hierarchical matrices based on a weak admissibility criterion
- Hierarchical matrices: algorithms and analysis
- Improved analysis of the subsampled randomized Hadamard transform
- Interpolation of spatial data. Some theory for kriging
- Nonstationary covariance models for global data
- ON STATIONARY PROCESSES IN THE PLANE
- On the Compression of Low Rank Matrices
- On the numerical solution of two-point boundary value problems
- On the numerical solution of two‐point boundary value problems II
- Parameter estimation in high dimensional Gaussian distributions
- Stochastic approximation of score functions for Gaussian processes
- Strictly and non-strictly positive definite functions on spheres
Cited in
(29)- Fast sampling of parameterised Gaussian random fields
- Efficient Fourier representations of families of Gaussian processes
- Equispaced Fourier representations for efficient Gaussian process regression from a billion data points
- A Scalable Method to Exploit Screening in Gaussian Process Models with Noise
- smashGP: Large-Scale Spatial Modeling via Matrix-Free Gaussian Processes
- FMM-LU: A Fast Direct Solver for Multiscale Boundary Integral Equations in Three Dimensions
- Efficient reduced-rank methods for Gaussian processes with eigenfunction expansions
- Linear-Cost Covariance Functions for Gaussian Random Fields
- A recursive skeletonization factorization based on strong admissibility
- Interpolative decomposition via proxy points for kernel matrices
- Tucker tensor analysis of Matérn functions in spatial statistics
- Scalable computations for nonstationary Gaussian processes
- Fast machine-precision spectral likelihoods for stationary time series
- Sparse Cholesky factorization for solving nonlinear PDEs via Gaussian processes
- Scalable Physics-Based Maximum Likelihood Estimation Using Hierarchical Matrices
- Dirichlet-Neumann averaging: the DNA of efficient Gaussian process simulation
- Scalable Gaussian Process Computations Using Hierarchical Matrices
- Parallel Skeletonization for Integral Equations in Evolving Multiply-Connected Domains
- Gaussian process learning via Fisher scoring of Vecchia's approximation
- On the maximum likelihood training of gradient-enhanced spatial Gaussian processes
- An adjoint for likelihood maximization
- A general framework for Vecchia approximations of Gaussian processes
- Estimating a matrix's singular values with interpolative decompositions
- Fast Coulomb matrix construction via compressing the interactions between continuous charge distributions
- Fast kernel summation in high dimensions via slicing and Fourier transforms
- Gaussian process regression with log-linear scaling for common non-stationary kernels
- Stochastic approximation of score functions for Gaussian processes
- Fitting Matérn smoothness parameters using automatic differentiation
- Kryging: geostatistical analysis of large-scale datasets using Krylov subspace methods
This page was built for publication: Fast spatial Gaussian process maximum likelihood estimation via skeletonization factorizations
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4601605)
