Convergence theory for the exact interpolation scheme with approximation vector as the first column of the prolongator and Rayleigh quotient iteration nonlinear smoother. (Q2360412)

This paper is concerned with a convergence analysis of the nonlinear two-level method with a nonlinear Rayleigh quotient iteration smoother, applied to the partial eigenvalue problem of the following article by \textit{R. Kužel} and \textit{P. Vaněk} [Numer. Linear Algebra Appl. 22, No. 6, 950--964 (2015; Zbl 1374.65203)]. The matrix is assumed to be symmetric and positive definite with a simple minimal eigenvalue. The authors seek the minimum eigenvalue and the corresponding eigenvector. The paper is organized as follows. Section 1 is an Introduction. In Section 2, the algorithm and the convergence result of the following article are presented -- [\textit{P. Fraňková, M. Hanuš, H. Kopincová, R. Kužel, I. Marek, I. Pultarová, P. Vaněk,} and \textit{Z. Vastl}, ``Convergence theory for the exact interpolation scheme with approximation vector as the first column of the prolongator: the partial eigenvalue problem'' (submitted to Numer. Math.)]. In the key Section 3, the nonlinear speed up of the Rayleigh quotient iteration smoother compared to the linear inverse power method is analyzed. Section 4 contains the final convergence theorem. The numerical tests with tables and graphical illustrations are presented in the final Section 5.