Multi-index stochastic collocation convergence rates for random PDEs with parametric regularity (Q506615)

The authors of this interesting paper concentrate on so-called multi-index stochastic collocation (MISC) methods to the approximation of an output from the solutions of linear elliptic partial differential equations of the form \[ -\operatorname{div}(x(x,y)\nabla u(x,y))=\zeta(x)\text{ in B }, \quad u(x,y)=0\text{ on } \partial B, \] where \(B\subset \mathbb R^d\) denotes the ``physical domain''. Section 2 introduces suitable assumptions and a class of random diffusion coefficients used throughout the paper. The MISC method is reviewed in Section 3. A complexity analysis of MISC with an infinite number of random variables is carried out in Section 4. A general convergence theorem is stated and proved. In Section 5, an application of MISC to the specific class of diffusion coefficients is discussed. Dependence of the convergence rate on the regularity of the diffusion coefficients is tracked. Results of extensive numerical experiments are summarized in Section 6. Finally, Section 7 provides some conclusions and final remarks. A technical appendix summarizes some technical results on the summability and regularity properties of certain random fields written in terms of their expansions.