Constructing continuous stationary covariances as limits of the second-order stochastic difference equations

In Bayesian statistical inverse problems the a priori probability distributions are often given as stochastic difference equations. We derive a certain class of stochastic partial difference equations by starting from second-order stochastic partial differential equations in one and two dimensions. We discuss discretisation schemes on uniform lattices of these stationary continuous-time stochastic processes and convergence of the discrete-time processes to the continuous-time processes. A special emphasis is given to an analytical calculation of the covariance kernels of the processes. We find a representation for the covariance kernels in a simple parametric form with controllable parameters: correlation length and variance. In the discrete-time processes the discretisation step is also given as a parameter. Therefore, the discrete-time covariances can be considered as discretisation-invariant. In the two-dimensional cases we find rotation-invariant and anisotropic representations of the difference equations and the corresponding continuous-time covariance kernels.

[1]  Sari Lasanen,et al.  Non-Gaussian statistical inverse problems. Part I: Posterior distributions , 2012 .

[2]  S. Varadhan,et al.  Diffusion processes with continuous coefficients, I , 1969 .

[3]  Tapio Helin,et al.  On infinite-dimensional hierarchical probability models in statistical inverse problems , 2009, 0907.5322.

[4]  Kiyosi Itô.  Stochastic Differential Equations , 2018, The Control Systems Handbook.

[5]  Didier Lucor,et al.  Covariance kernel representations of multidimensional second-order stochastic processes , 2006, J. Comput. Phys..

[6]  Erkki Somersalo,et al.  Linear inverse problems for generalised random variables , 1989 .

[7]  Matti Lassas. Eero Saksman,et al.  Discretization-invariant Bayesian inversion and Besov space priors , 2009, 0901.4220.

[8]  Bernt Øksendal,et al.  WHITE NOISE. AN INFINITE DIMENSIONAL CALCULUS , 1995 .

[9]  J. Miller.  Numerical Analysis , 1966, Nature.

[10]  S. Siltanen,et al.  Can one use total variation prior for edge-preserving Bayesian inversion? , 2004 .

[11]  J. Doob.  Stochastic processes , 1953 .

[12]  Markku Lehtinen,et al.  Fortran linear inverse problem solver , 2010 .

[13]  V. Bogachev,et al.  Open mappings of probability measures and the skorokhod representation theorem , 2002 .

[14]  I. S. Gradshteyn,et al.  Table of Integrals, Series, and Products , 1976 .

[15]  Donald E. Knuth.  Two notes on notation , 1992 .

[16]  Kenneth E. Iverson,et al.  A programming language , 1899, AIEE-IRE '62 (Spring).

[17]  Leonhard Held,et al.  Gaussian Markov Random Fields: Theory and Applications , 2005 .

[18]  Hui-Hsiung Kuo,et al.  White noise distribution theory , 1996 .

[19]  E. Somersalo,et al.  Statistical and computational inverse problems , 2004 .

[20]  Amiel Feinstein,et al.  Applications of harmonic analysis , 1964 .

[21]  S. Varadhan,et al.  Diffusion processes with boundary conditions , 1971 .

[22]  H. Rue,et al.  An explicit link between Gaussian fields and Gaussian Markov random fields: the stochastic partial differential equation approach , 2011 .

[23]  G. Winkler,et al.  The Stochastic Integral , 1990 .