A symplectic acceleration method for the solution of the algebraic Riccati equation on a parallel computer (Q1260795)

The purpose of this paper is to describe an effective acceleration method which can be used to speed up the rate of convergence of the symplectic Jacobi-like algorithms. The starting point is the algebraic Riccati equation. A brief introduction to Householder and Jacobi symplectic rotations is given, reordering techniques for eigenvalues and its parallel implementations are proposed. The next step is to derive a cubic symplectic acceleration method and its parallel processing. It follows a quantitative convergence analysis of this method. It is proved that the method converges cubically. The authors compare this method with the symplectic Jacobi-like method of \textit{R. Byers} [A Hamiltonian-Jacobi algorithm, presented at SIAM conference on Control in the '90s, May 1989]. Some examples are computed by the last one and then corrected by the symplectic acceleration method described here. You can find informations about the flop counts in both methods and some numerical results.