Marian Brezina, CU-Boulder; Rob Falgout, CASC, LLNL; Tom Manteuffel, CU-Boulder; Steve McCormick, CU-Boulder; John Ruge, CU-Boulder
Substantial effort has recently focused on developing methods capable of solving very large linear systems that arise from discretizing partial differential equations, especially on unstructured grids. Algebraic multigrid (AMG) is of particular interest because of its promise of optimal performance without the need for explicit knowledge of the problem's origin. We introduce an extension of AMG based on an adaptive process that achieves good convergence on a broader class of problems than the original algorithm.