Inexact inverse iteration for symmetric matrices (Q2496620)

For large and sparse symmetric eigenvalue problem, an important method is the inner-outer iterative method: the outer iteration is the basic inverse iteration algorithm, with the inner iteration being the inexact solution of shifted linear systems. The authors discuss the effect of inexact solvers. A convergence theory independent of the details of the inexact iterative solver is presented. New bounds on the number of inner iterations at each step of the outer iteration are proved. Some other interesting results on preconditioned Rayleigh quotient iteration are given. A key feature of this paper is that a residual stopping condition is widely used.