Krylov methods for solving linear systems

Krylov methods for solving linear systems are generally used with a preconditioner which accelerates the convergence. They only require matrix multiplication by a vector, scalar products and vector additions. This article explains these methods and their different aims. An in-depth analysis of the Krylov methods is then provided: construction of the basis, GMRES and FOM methods, conjugate gradient, BiCG and BiCGstab or QMR methods. An example of methods concludes this article.