Breakdowns and stagnation in iterative methods (Q1359974)

The paper deals with iterative methods based on the use of Krylov subspaces for linear systems \(Au=f\). The author analyzes different kinds of breakdown and shows that one of them is related to the choice of the basis vectors and can be avoided by passing to the so-called orthodir basis vectors. The kind of breakdown, associated with finding the coefficients, is studied as well. The stagnation in the case of the BiCG method and its variants is considered; sometimes it can even imply breakdown. It should be noted that in solving large systems, some other modern methods, based on preconditioning and symmetrization, can be helpful [see, e.g., \textit{E. G. D'yakonov}, Optimization in solving elliptic problems (1996; Zbl 0852.65087)].