大規模最適化問題、グラフ探索、機械学習やデジタルツインなど

旧名:最適化問題に対する超高速&安定計算

Scalable Single Source Shortest Path Algorithms for Massively Parallel Systems

以下の論文が IPDPS 2014 で発表されました。個人的には一番面白かった発表です。

Scalable Single Source Shortest Path Algorithms forMassively Parallel Systems

Venkatesan T. Chakaravarthy, Fabio Checconi, Fabrizio Petrini, Yogish Sabharwal

Abstract―In the single-source shortest path (SSSP) problem,we have to find the shortest paths from a source vertex v to all
other vertices in a graph. In this paper, we introduce a novel parallel algorithm, derived from the Bellman-Ford and Deltastepping algorithms. We employ various pruning techniques, such as edge classification and direction-optimization, to dramatically
reduce inter-node communication traffic, and we propose load balancing strategies to handle higher-degree vertices. The extensive performance analysis shows that our algorithms work well on scale-free and real-world graphs. In the largest tested configuration, an R-MAT graph with 2^38 vertices and 2^42 edgeson 32,768 Blue Gene/Q nodes, we have achieved a processing rate of three Trillion Edges Per Second (TTEPS), a four orders of magnitude improvement over the best published results.