An orthogonal accelerated deflation technique for large symmetric eigenproblems (Q1192738)

Eigenvalue algorithms for large sparse symmetric generalized matrix eigenvalue problems based on conjugate gradient minimization of the Rayleigh quotient with incomplete Chebyshev preconditioning are discussed. It is shown on a numerical example that it is faster to use successive deflation than simultaneous iteration, when several of the smallest eigenvalues are sought, (at least when these are simple).