Polynomial Iteration for Nonsymmetric Indefinite Linear Systems.

摘要

We examine iterative methods for solving sparse nonsymmetric indefinite systems of linear equations. Methods considered include a adaptive model based on polynomials that satisfy an optimality condition in the Chebyshev norm, the conjugate gradient-like method GMRES, and the conjugate gradient method applied to the normal equations. Numerical experiments on several non-self-adjoint indefinite elliptic boundary value problems suggest that none of these methods is dramatically superior to the others. Their performance in solving moderately difficult problems is satisfactory, but for harder problems their convergence is slow.