Efficient Algebraic Multigrid Algorithms and Their Convergence

From MaRDI portal





The authors present 8 new interpolation formulae for use in algebraic multigrid algorithms. For this aim, they separate suitably defined strongly and weakly connected unknowns and exploit Gauss-Seidel or Jacoby iteration-type ideas. In the scetch of a convergence theory, the authors refer strongly to earlier work by \textit{J. W. Ruge} and \textit{K. Stüben} [Algebraic multigrid methods. S. F. McCormick (ed.). Frontiers in Applied Mathematics, 3. SIAM, 73-130 (1987; Zbl 0659.65094)], and by \textit{W. Huang} [Appl. Math. Comput. 46, 145-164 (1991; Zbl 0743.65029)], and assume symmetric positive definite matrices with weakly dominant diagonal. Finally, they extensively report on numerical experiments on 10 different problems (including 2D and 3D Poisson equations, the 2D linear elasticity and the biharmonic equations, as well as several anisotropic and related problems which are known to be difficult for classical algebraic multigrid). However, comparison to methods of other authors is not given. The reported convergence rates are (if not counting the biharmonic case) surprisingly good but shown for two different grid sizes only. In the case of the elasticity system the memory needed for all matrices in some of the proposed methods is exceeding that for the finest grid matrix by a factor of 16 (which factor is, for the other problems, near to 2).




Cited in
(31)








This page was built for publication: Efficient Algebraic Multigrid Algorithms and Their Convergence

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4785937)