Augmented Householder Eigenvalue Solver

Computes a few eigenvalues and eigenvectors of a nonsymmetric matrix.

AHBEIGS: will find a few eigenvalues and eigenvectors for either the standard eigenvalue problem A*x = lambda*x or the generalized eigenvalue problem A*x = lambda*B*x.

[V,D,FLAG] = AHBEIGS(A,OPTIONS)

[V,D,FLAG] = AHBEIGS(A,B,OPTIONS)

[V,D,FLAG] = AHBEIGS('Afunc',N,B,OPTIONS)

The first input argument must be a matrix A which can be passed as a numeric matrix or as a M-file ('Afunc') that computes the product A*X or inv(B)*A*X, where X is a (N x blsz) matrix.

The program is based on a block Arnoldi method, that makes use of Householder reflections to maintain orthogonality and restarting is accomplished by augmentation of the Krylov subspace with Schur vectors.

Research supported by NSF grant DMS-0311786.