scientific article; zbMATH DE number 4020318

From MaRDI portal
Publication:3763472





[For the entire collection see Zbl 0623.00017.] The Lanczos algorithm, originally devised to tridiagonalize a matrix, is used for the generalized eigenvalue problem to operate on a whole subspace, yielding a block-tridiagonal matrix in the Krylov sequence of subspaces. To use prior information and enhance convergence, a subspace restart formulation is introduced. Global orthogonality of the iterated vectors is maintained by a double Gram-Schmidt re-orthogonalization.











This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3763472)