Robust Preconditioners for Saddle Point Problems

We survey preconditioning methods for matrices on saddle point form, as typically arising in constrained optimization problems. Special consideration is given to indefinite matrix preconditioners and a preconditioner which results in a symmetric positive definite matrix, which latter may enable the use of the standard conjugate gradient (CG) method. These methods result in eigenvalues with positive real parts and small or zero imaginary parts. The behaviour of some of these techniques is illustrated on solving a regularized Stokes problem.