On a variable smoothing procedure for Krylov subspace methods (Q1375088)

The convergence behavior of the Galerkin-Krylov subspace method for solving linear systems can be very erratic, and therefore a smoothing technique or a minimal residual seminorm variant of these Galerkin methods can be proposed to eliminate this problem. The authors examine a class of minimal residual seminorm methods, and show that these methods can be obtained from a variable smoothing technique applied to Galerkin methods.