Conjugate gradient methods for continuation problems. II (Q1917822): Difference between revisions

[For part I see \textit{E. L. Allgower}, \textit{C.-S. Chien}, \textit{K. Georg} and \textit{C.-F. Wang}, ibid. 38, No. 1-3, 1-16 (1991; Zbl 0753.65046).] The paper is concerned with a study of the linear conjugate gradient method for large sparse continuation problems. This method is incorporated as linear solver in the context of a continuation method. The branches of the solution of linear and nonlinear eigenvalue problems are approximated using the generalized minimal residual (GMRES) algorithm of \textit{Y. Saad} and \textit{M. H. Schultz} [SIAM J. Sci. Stat. Comput. 7, 856-869 (1986; Zbl 0599.65018)]. It is also indicated how this preconditioned algorithm may by used to test simple and multiple bifurcations without solving the bordered linear systems. An efficient higher-order predictor-Newton corrector confirmation method is proposed with the implementation based on the GMRES algorithm. How to use a special conjugate gradient method to perform the corrector phase is also described. In both cases the investigation deals with the perturbation problems for the bifurcation. Samples of numerical results concerning certain nonlinear eigenvalue problems and concluding remarks concerning the performance of three-term conjugate gradient methods are also given.