Tensor-sparsity of solutions to high-dimensional elliptic partial differential equations (Q330094)

The paper delas with the solution of problems having a large number of spatial variables where the use of standard methods leads to exponential increase of the computational complexity. One possibility is to employ some form of sparse tensor approximation. The main purpose of the paper is to propose specific notions of sparsity, based on tensor decompositions, and then show that the solutions of certain high-dimensional diffusion equations inherit this type of sparsity from given data. This is then shown to lead indeed to tractability of solving such high-dimensional partial differential equations. The authors present a Scheme-Exp numerical algorithm and formulate a theorem, which bounds the number of computations necessary for achieving a prescribed accuracy.