par Napov, Artem ;Notay, Yvan
Référence Electronic transactions on numerical analysis, 45, page (201-218)
Publication Publié, 2016
Article révisé par les pairs
Résumé : We consider linear systems whose matrices are Laplacians of undirected graphs. We present a new aggregation-based algebraic multigrid method designed to achieve robustness for this class of problems, despite the diversity of connectivity patterns encountered in practical applications. These indeed range from regular mesh graphs to scale-free type of graphs associated with social networks. The method is based on the recursive static elimination of the vertices of degree 1 combined with a new Degree-aware Rooted Aggregation (DRA) algorithm. This algorithm always produces aggregates big enough so that the cost per iteration is low, whereas reasonable convergence is observed when the approach is combined with the K-cycle. The robustness of the resulting method is illustrated on a large collection of test problems, and its effectiveness is assessed via the comparison with a state-of-the-art reference method.