On theoretical and numerical aspects of symplectic Gram-Schmidt-like algorithms (Q2501237)

Symplectic Gram-Schmidt-like algorithms play an important role in the numerical solution of large sparse Hamiltonian matrix eigenvalue problems. This paper examines, both theoretically and by numerical experiment, the extent to which the numerical properties of these algorithms can be improved by modifications in implementation. The results are compared with the work of \textit{A. Björck} [BIT 7, 1--21 (1967; Zbl 0183.17802)] on the classical Gram-Schmidt algorithm.