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



Similar documents
Visualizing of Berkeley Earth, NASA GISS, and Hadley CRU averaging techniques

Spatial Statistics Chapter 3 Basics of areal data and areal data modeling

BayesX - Software for Bayesian Inference in Structured Additive Regression

Statistics Graduate Courses

An Introduction to Machine Learning

Exact Inference for Gaussian Process Regression in case of Big Data with the Cartesian Product Structure

THE NUMBER OF GRAPHS AND A RANDOM GRAPH WITH A GIVEN DEGREE SEQUENCE. Alexander Barvinok

Gaussian Conjugate Prior Cheat Sheet

Two Topics in Parametric Integration Applied to Stochastic Simulation in Industrial Engineering

Introduction to the Finite Element Method

Machine Learning and Data Analysis overview. Department of Cybernetics, Czech Technical University in Prague.

Model-based Synthesis. Tony O Hagan

MATH 425, PRACTICE FINAL EXAM SOLUTIONS.

Modeling and Performance Evaluation of Computer Systems Security Operation 1

Confidence Intervals for the Difference Between Two Means

Cell Phone based Activity Detection using Markov Logic Network

ANALYZING INVESTMENT RETURN OF ASSET PORTFOLIOS WITH MULTIVARIATE ORNSTEIN-UHLENBECK PROCESSES

Bandwidth Selection for Nonparametric Distribution Estimation

Modern Optimization Methods for Big Data Problems MATH11146 The University of Edinburgh

Natural cubic splines

Introduction to General and Generalized Linear Models

Moving Least Squares Approximation

Course: Model, Learning, and Inference: Lecture 5

Statistical Analysis of Non-stationary Waves off the Savannah Coast, Georgia, USA

Methods of Data Analysis Working with probability distributions

THE DYING FIBONACCI TREE. 1. Introduction. Consider a tree with two types of nodes, say A and B, and the following properties:

Linear Threshold Units

LECTURE 4. Last time: Lecture outline

SPECTRAL POLYNOMIAL ALGORITHMS FOR COMPUTING BI-DIAGONAL REPRESENTATIONS FOR PHASE TYPE DISTRIBUTIONS AND MATRIX-EXPONENTIAL DISTRIBUTIONS

INTRODUCTION TO GEOSTATISTICS And VARIOGRAM ANALYSIS

Using the Theory of Reals in. Analyzing Continuous and Hybrid Systems

Understanding and Applying Kalman Filtering

Part 2: Analysis of Relationship Between Two Variables

11. Time series and dynamic linear models

Matrices and Polynomials

OUTLIER ANALYSIS. Data Mining 1

Extrinsic geometric flows

Finitely Additive Dynamic Programming and Stochastic Games. Bill Sudderth University of Minnesota

BINOMIAL OPTIONS PRICING MODEL. Mark Ioffe. Abstract

Market Risk Analysis. Quantitative Methods in Finance. Volume I. The Wiley Finance Series

SOLVING LINEAR SYSTEMS

MULTIVARIATE PROBABILITY DISTRIBUTIONS

Analysis and its Applications. A unique opportunity for fully funded PhD study in Edinburgh in all areas of analysis and its applications.

Some probability and statistics

INDIRECT INFERENCE (prepared for: The New Palgrave Dictionary of Economics, Second Edition)

CITY UNIVERSITY LONDON. BEng Degree in Computer Systems Engineering Part II BSc Degree in Computer Systems Engineering Part III PART 2 EXAMINATION

Tutorial on Markov Chain Monte Carlo

Probability for Estimation (review)

Exploratory Data Analysis

U-statistics on network-structured data with kernels of degree larger than one

Univariate and Multivariate Methods PEARSON. Addison Wesley

Approximate Likelihoods for Spatial Processes

Geostatistics Exploratory Analysis

Department of Mathematics, Indian Institute of Technology, Kharagpur Assignment 2-3, Probability and Statistics, March Due:-March 25, 2015.

Monte Carlo and Empirical Methods for Stochastic Inference (MASM11/FMS091)

Lecture 3: Linear methods for classification

Binary Image Reconstruction

Inner Product Spaces

FINITE DIFFERENCE METHODS

Introduction to Support Vector Machines. Colin Campbell, Bristol University

Computational Optical Imaging - Optique Numerique. -- Deconvolution --

Classification Problems

VERTICES OF GIVEN DEGREE IN SERIES-PARALLEL GRAPHS

1. (First passage/hitting times/gambler s ruin problem:) Suppose that X has a discrete state space and let i be a fixed state. Let

Machine Learning in Statistical Arbitrage

AN INTRODUCTION TO NUMERICAL METHODS AND ANALYSIS

Probabilistic Models for Big Data. Alex Davies and Roger Frigola University of Cambridge 13th February 2014

Data Mining: Algorithms and Applications Matrix Math Review

CONTROL SYSTEMS, ROBOTICS, AND AUTOMATION - Vol. V - Relations Between Time Domain and Frequency Domain Prediction Error Methods - Tomas McKelvey

Simple Linear Regression Inference

POISSON AND LAPLACE EQUATIONS. Charles R. O Neill. School of Mechanical and Aerospace Engineering. Oklahoma State University. Stillwater, OK 74078

Probability and Random Variables. Generation of random variables (r.v.)

Statistics & Probability PhD Research. 15th November 2014

Finite Differences Schemes for Pricing of European and American Options

Linear Classification. Volker Tresp Summer 2015

Time Series Analysis

Using the Delta Method to Construct Confidence Intervals for Predicted Probabilities, Rates, and Discrete Changes

MATH 10: Elementary Statistics and Probability Chapter 5: Continuous Random Variables

Gaussian Processes to Speed up Hamiltonian Monte Carlo

Portfolio Distribution Modelling and Computation. Harry Zheng Department of Mathematics Imperial College

Impulse Response Functions

Support Vector Machines for Classification and Regression

Advanced Signal Processing and Digital Noise Reduction

1 VECTOR SPACES AND SUBSPACES

State Space Time Series Analysis

ANALYTICS IN BIG DATA ERA

A SURVEY ON CONTINUOUS ELLIPTICAL VECTOR DISTRIBUTIONS

Some remarks on two-asset options pricing and stochastic dependence of asset prices

How To Solve A Sequential Mca Problem

Pricing Barrier Option Using Finite Difference Method and MonteCarlo Simulation

An Additive Neumann-Neumann Method for Mortar Finite Element for 4th Order Problems

Transcription:

Intro B, W, M, & R SPDE/GMRF Example End An explicit link between Gaussian fields and Gaussian Markov random fields; the stochastic partial differential equation approach Finn Lindgren 1 Håvard Rue 1 Johan Lindström 2 1 Department of Mathematical Sciences Norwegian University of Science and Technology, Trondheim, Norway 2 Mathematical Statistics Centre for Mathematical Sciences, Lund University, Lund, Sweden RSS, London, 16th March 2011

Intro B, W, M, & R SPDE/GMRF Example End Efficiency Gaussian Markov random fields Computationally efficent spatial models Spatial random field models Dense covariance: x N (µ, Σ) = expensive computations Sparse precision: x N (µ, Q 1 ) = efficient computations Leukaemia survival in NW England

Intro B, W, M, & R SPDE/GMRF Example End Efficiency Gaussian Markov random fields Irregular discrete data locations lattice CAR models We need models with well-defined continuous interpretation

Intro B, W, M, & R SPDE/GMRF Example End Efficiency Gaussian Markov random fields Irregular discrete data locations triangular graph We need models with well-defined continuous interpretation

Intro B, W, M, & R SPDE/GMRF Example End Efficiency Gaussian Markov random fields Irregular discrete data locations refined triangulation We need models with well-defined continuous interpretation

Intro B, W, M, & R SPDE/GMRF Example End Efficiency Gaussian Markov random fields Irregular discrete data locations smooth triangulation We need models with well-defined continuous interpretation

Intro B, W, M, & R SPDE/GMRF Example End Efficiency Gaussian Markov random fields Irregular discrete data locations continuous space We need models with well-defined continuous interpretation

Intro B, W, M, & R SPDE/GMRF Example End CAR Matérn Markov Whittle Conditional autoregression models on lattices GMRF Covariance CAR(2) κ u K 1 (κ u ) Whittle (1954) CAR(1) 1 2π K 0(κ u ) Besag (1981) ICAR(1) 1 2π log( u ) Besag & Mondal (2005)

Intro B, W, M, & R SPDE/GMRF Example End CAR Matérn Markov Whittle Matérn covariances (Bertil Matérn) The Matérn covariance family Cov(x(0), x(u)) = σ 2 21 ν Γ(ν) (κ u )ν K ν (κ u ), u R d Scale κ > 0, smoothness ν > 0, variance σ 2 > 0 Are any other discrete domain Markov fields close to Matérn fields? Are any Matérn fields continuous domain Markov fields?

Intro B, W, M, & R SPDE/GMRF Example End CAR Matérn Markov Whittle The continuous domain Markov property S is a separating set for A and B: x(a) x(b) x(s) A S B

Intro B, W, M, & R SPDE/GMRF Example End CAR Matérn Markov Whittle Identifying a Markov field by its spectrum Spectral representation: x(u) = exp(iu k) x(k) dk R d Spectrum: S x (k) = E( x(k) 2 ) Rozanov (1977) x(u) is Markov S x (k) P(k) 1 where P(k) is a non-negative symmetric polynomial What is the spectrum of a Matérn field? c MFO

Intro B, W, M, & R SPDE/GMRF Example End CAR Matérn Markov Whittle The Whittle-Matérn connection Whittle (1954, 1963) Matérn fields on R d are stationary solutions to the SPDE ( κ 2 ) α/2 x(u) = W(u), α = ν + d/2 The spectrum is S x (k) = (2π) d (κ 2 + k 2 ) α Therefore Matérn fields are Markov when ν + d/2 is an integer. Laplacian: = d 2 u 2 i=1 i

Intro B, W, M, & R SPDE/GMRF Example End Finite Projection Markov Lattices... and beyond Piecewise linear representations x(u) = cos(u 1 ) + sin(u 2 ) x(u) = k ψ k(u) x k Guiding principle Attack the SPDE with local finite dimensional representations instead of covariances or kernels (subsets of Green s functions)!

Intro B, W, M, & R SPDE/GMRF Example End Finite Projection Markov Lattices... and beyond The best piecewise linear approximation k ψ k(u) x k Projection of the SPDE: Linear systems of equations (α = 2) j (κ 2 ψ i, ψ j + ψ i, ψ j } {{ } } {{ } C ij G ij )x j D = ψ i, W jointly for all i. C and G are as sparse as the triangulation neighbourhood Constructing the precision matrices K = κ 2 C + G α = 1 α = 2 α = 3, 4,... Kx N (0, K) N (0, C) ( ) N 0, CQ 1 x,α 2 C Q x,α K K T C 1 K K T C 1 Q x,α 2 C 1 K

Intro B, W, M, & R SPDE/GMRF Example End Finite Projection Markov Lattices... and beyond Markov approximation C 1 is dense, which leads to non-markov models! The diagonal matrix C, C i,i = ψ i, 1 = the local cell area, gives a sparse precision with the total energy preserved.

Intro B, W, M, & R SPDE/GMRF Example End Finite Projection Markov Lattices... and beyond Lattice on R 2, node distance h, regular triangulation Order α = 1: generalised covariance 1 2π K 0(κ u ) 1 κ 2 h 2 1 + 1 4 1 1 Order α = 2: covariance 1 4πκ 2 κ u K 1 (κ u ) κ 4 h 2 1 1 + 2κ 2 1 4 1 + 1 h 1 2 1 2 8 2 1 8 20 8 1 2 8 2 1

Intro B, W, M, & R SPDE/GMRF Example End Finite Projection Markov Lattices... and beyond Beyond classical Matérn models The approach can in a straightforward way be extended to oscillating, anisotropic, non-stationary, non-separable spatio-temporal, and multivariate fields on manifolds. (κ 2 ) α/2 (τ x(u)) = W(u), u R d

Intro B, W, M, & R SPDE/GMRF Example End Finite Projection Markov Lattices... and beyond Beyond classical Matérn models The approach can in a straightforward way be extended to oscillating, anisotropic, non-stationary, non-separable spatio-temporal, and multivariate fields on manifolds. (κ 2 e iπθ ) α/2 (τ x(u)) = W(u), u R d

Intro B, W, M, & R SPDE/GMRF Example End Finite Projection Markov Lattices... and beyond Beyond classical Matérn models The approach can in a straightforward way be extended to oscillating, anisotropic, non-stationary, non-separable spatio-temporal, and multivariate fields on manifolds. (κ 2 + m M ) α/2 (τx(u)) = W(u), u R d

Intro B, W, M, & R SPDE/GMRF Example End Finite Projection Markov Lattices... and beyond Beyond classical Matérn models The approach can in a straightforward way be extended to oscillating, anisotropic, non-stationary, non-separable spatio-temporal, and multivariate fields on manifolds. (κ 2 u + m u M u ) α/2 (τ u x(u)) = W(u), u R d

Intro B, W, M, & R SPDE/GMRF Example End Finite Projection Markov Lattices... and beyond Beyond classical Matérn models The approach can in a straightforward way be extended to oscillating, anisotropic, non-stationary, non-separable spatio-temporal, and multivariate fields on manifolds. (κ 2 u + m u M u ) α/2 (τ u x(u)) = W(u), u Ω

Intro B, W, M, & R SPDE/GMRF Example End Finite Projection Markov Lattices... and beyond Beyond classical Matérn models The approach can in a straightforward way be extended to oscillating, anisotropic, non-stationary, non-separable spatio-temporal, and multivariate fields on manifolds. (κ 2 u + m u M u ) α/2 (τ u x(u)) = W(u), u Ω

Intro B, W, M, & R SPDE/GMRF Example End Finite Projection Markov Lattices... and beyond Beyond classical Matérn models The approach can in a straightforward way be extended to oscillating, anisotropic, non-stationary, non-separable spatio-temporal, and multivariate fields on manifolds. (κ 2 u + m u M u ) α/2 (τ u x(u)) = W(u), u Ω

Intro B, W, M, & R SPDE/GMRF Example End Finite Projection Markov Lattices... and beyond Beyond classical Matérn models The approach can in a straightforward way be extended to oscillating, anisotropic, non-stationary, non-separable spatio-temporal, and multivariate fields on manifolds. ( t + κ2 u,t + m u,t M u,t ) (τ u,t x(u, t)) = E(u, t), (u, t) Ω R

Intro B, W, M, & R SPDE/GMRF Example End Large problems GHCN INLA Climate Weather Large spatial problems: covariances are impractical Data points n 2500 20 Kriging locations m 15000 20 Estimation: time O(n 3 ), storage O(n 2 ) 20Gbytes Kriging: time O(mn + n 3 ), storage O(mn + n 2 ) 130Gbytes This is not a huge data set.

Intro B, W, M, & R SPDE/GMRF Example End Large problems GHCN INLA Climate Weather Practical triangulation based GMRFs Data points n 2500 20 Basis functions m 15000 20 Estimation+Kriging: time O(m 3/2 ), storage O(m + n) 50Mbytes

Intro B, W, M, & R SPDE/GMRF Example End Large problems GHCN INLA Climate Weather Global temperature analysis http://www.ncdc.noaa.gov/ghcn/ghcn.html Monthly average temperatures On average 2500 locations per year Covariate data (elevation, land use, etc.) for 7000 stations Model for yearly mean temperatures Observations : temperature elevation + climate + anomaly Climate : µ u, τ u, κ 2 u = k ψ k (u)θ k Anomaly : ( κ 2 u ) (τ u x t (u)) = W(u) Weather : µ(u) + x t (u)

Intro B, W, M, & R SPDE/GMRF Example End Large problems GHCN INLA Climate Weather Code outline for Bayesian inference with R-INLA mesh = inla.mesh(loc=station.locations, ## optional cutoff=10/earth.radius, refine=list(max.edge=500/earth.radius))... spde = inla.spde(mesh, model="matern",... param=list(basis.t=b.tau,...))... formula = temperature ~ elevation + climate.mu + f(anomaly, model=spde, replicate=year)... result = inla(formula, data, family="gaussian",...)

Intro B, W, M, & R SPDE/GMRF Example End Large problems GHCN INLA Climate Weather Estimated climate covariance properties 1970-1989 Approximate range Standard deviations km 0 1000 2000 3000 C 0.0 1.0 2.0 3.0 90 60 30 0 30 60 90 Latitude 90 60 30 0 30 60 90 Latitude

Intro B, W, M, & R SPDE/GMRF Example End Large problems GHCN INLA Climate Weather Empirical Climate 1970 1989 (C) 30 20 10 0 10 20 30 Latitude 90 60 45 30 15 0 15 30 45 60 90 180 135 90 45 0 45 90 135 180

Intro B, W, M, & R SPDE/GMRF Example End Large problems GHCN INLA Climate Weather Std dev for Climate 1970 1989 (C) 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 Latitude 90 60 45 30 15 0 15 30 45 60 90 180 135 90 45 0 45 90 135 180

Intro B, W, M, & R SPDE/GMRF Example End Large problems GHCN INLA Climate Weather Empirical Anomaly 1980 (C) 2.0 1.5 1.0 0.5 0.0 0.5 1.0 1.5 2.0 Latitude 90 60 45 30 15 0 15 30 45 60 90 180 135 90 45 0 45 90 135 180

Intro B, W, M, & R SPDE/GMRF Example End Large problems GHCN INLA Climate Weather Std dev for Anomaly 1980 (C) 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 Latitude 90 60 45 30 15 0 15 30 45 60 90 180 135 90 45 0 45 90 135 180

Intro B, W, M, & R SPDE/GMRF Example End Conclusion SPDE/GMRF The connection allows for consistent local specifications of general spatio-temporal models that allow efficient practical computations. Unavoidably beyond the traditional computational statistics toolbox Combines standard concepts from different areas in applied mathematics and engineering Detailed knowledge of SPDEs is not needed for practical use Open source software available: http://www.r-inla.org/

Intro B, W, M, & R SPDE/GMRF Example End Covariances for Beyond classical Matérn models The approach can in a straightforward way be extended to oscillating, anisotropic, non-stationary, non-separable spatio-temporal, and multivariate fields on manifolds. (κ 2 ) α/2 (τ x(u)) = W(u), u R d

Intro B, W, M, & R SPDE/GMRF Example End Covariances for Beyond classical Matérn models The approach can in a straightforward way be extended to oscillating, anisotropic, non-stationary, non-separable spatio-temporal, and multivariate fields on manifolds. (κ 2 e iπθ ) α/2 (τ x(u)) = W(u), u R d

Intro B, W, M, & R SPDE/GMRF Example End Covariances for Beyond classical Matérn models The approach can in a straightforward way be extended to oscillating, anisotropic, non-stationary, non-separable spatio-temporal, and multivariate fields on manifolds. (κ 2 + m M ) α/2 (τx(u)) = W(u), u R d

Intro B, W, M, & R SPDE/GMRF Example End Covariances for Beyond classical Matérn models The approach can in a straightforward way be extended to oscillating, anisotropic, non-stationary, non-separable spatio-temporal, and multivariate fields on manifolds. (κ 2 u + m u M u ) α/2 (τ u x(u)) = W(u), u R d

Intro B, W, M, & R SPDE/GMRF Example End Covariances for Beyond classical Matérn models The approach can in a straightforward way be extended to oscillating, anisotropic, non-stationary, non-separable spatio-temporal, and multivariate fields on manifolds. (κ 2 u + m u M u ) α/2 (τ u x(u)) = W(u), u Ω

Intro B, W, M, & R SPDE/GMRF Example End Covariances for Beyond classical Matérn models The approach can in a straightforward way be extended to oscillating, anisotropic, non-stationary, non-separable spatio-temporal, and multivariate fields on manifolds. (κ 2 u + m u M u ) α/2 (τ u x(u)) = W(u), u Ω

Intro B, W, M, & R SPDE/GMRF Example End Covariances for Beyond classical Matérn models The approach can in a straightforward way be extended to oscillating, anisotropic, non-stationary, non-separable spatio-temporal, and multivariate fields on manifolds. (κ 2 u + m u M u ) α/2 (τ u x(u)) = W(u), u Ω