A new characterization of random fields appearing in physical models ispresented that is based on their well-known Homogeneous Chaos expansions. Wetake advantage of the adaptation capabilities of these expansions where thecore idea is to rotate the basis of the underlying Gaussian Hilbert space, inorder to achieve reduced functional representations that concentrate theinduced probability measure in a lower dimensional subspace. For a smoothfamily of rotations along the domain of interest, the uncorrelated Gaussianinputs are transformed into a Gaussian process, thus introducing a mesoscalethat captures intermediate characteristics of the quantity of interest.
展开▼
