A Cholesky LR algorithm for the positive definite symmetric diagonal-plus-semiseparable eigenproblem (Q2465323)

Many algorithms are known for solving the symmetric eigenvalue problem. Among others, an orthogonal similarity reduction was recently presented to reduce a symmetric matrix into a diagonal-plus-semiseparable one (denoted DPSS) with free choice of the diagonal. This paper presents an efficient Cholesky LR algorithm to compute the eigenvalues of a positive definite DPSS matrix, preserving the DPSS structure. A useful feature is that the eigenvalues are computed from the smallest to the largest one, so this algorithm looks very suitable for those problems, where only a couple of the smallest eigenvalues are required. Reported numerical results show that the accuracy exhibited by the presented methods looks competitive with analogous LAPACK routines.