Flexible Krylov solvers for shifted systems with applications to solving large-scale inverse problems

Tania Bakhos (Stanford University)

Frank Adams 1,

Inverse problems involve estimating parameters from physical measurements; for example, hydraulic tomography is a method to estimate hydraulic parameters related to the subsurface from pressure measurements obtained by a series of pumping tests. For many inverse problems, the solution of a shifted system of linear equations is the major computational bottleneck. By exploiting the shift-invariance property of Krylov subspaces, only a single Krylov basis is computed and the solution for multiple shifts can be obtained at a cost that is similar to the cost of solving a single system. We have developed flexible Krylov solvers for shifted systems and proposed fast methods for solving large-scale inverse problems based on these Krylov solvers. We demonstrate the performance gains (up to 20x speedup) on data sets taken from hydraulic tomography.

Import this event to your Outlook calendar
▲ Up to the top