A refined subspace iteration algorithm for large sparse eigenproblems (Q1964385)

Two innovations to subspace iteration are presented. Such iterations, which are extensions of the power method, are very popular for computing a few dominant eigenvalues and eigenvectors of large sparse matrices. It is shown that the refined algorithms derived from the proposed innovations are significantly better than conventional algorithms if the dimension of the subspace is larger than the number of desired eigenpairs. Three numerical examples are given and the relative efficiency of the proposed algorithms is shown as a result of a set of computational experiments.