engleski

On Simulations of Random Fields on Bounded Domains

A random field on a bounded domain is usually estimated from data by using the linear least square estimation. To apply this method it is necessary to know the covariance function, while distribution of field at a point is not used. Better results of estimation are expected by applying methods for which the distribution of field must be utilized. We consider here a method for which the random field is represented by using the Loeve-Karhunen decomposition. Numerical approximations are obtained by grids discretizing the domain and by solving the corresponding finite-dimensional eigenvalue problem for finite dimensional matrices.

Random field; Monte Carlo methods

nije evidentirano