A GPU-Based Solution to Fast Calculation of Betweenness Centrality on Large Weighted Networks

2017-01-20T00:38:25Z (GMT) by Jichang Zhao
<div> <div> <p>We develop an efficient parallel GPU-based approach to boost the calculation of betweenness centrality on very large and weighted networks. Comprehensive and systematic evaluations on both synthetic and real-world networks demonstrate that our solution can arrive the performance of 30x to 150x speedup over the CPU implementation by integrating the work-efficient and warp-centric strategies. </p><p><br></p><p>README</p><p>This program compute node or edge betweenness centrality on weighted networks.</p><p><br></p><p>Graph format</p><p>The first line has two columns, the first item is the number of nodes and the second one is the number of edges. The following lines are weighted edges that contain three columns. The first two items are the node pair and the last item is the weight of the link. One edge appears only once, e.g., if "0 1 0.5" is in the file, the edge "1 0 0.5" must not show up. You can find an example file named "graph.txt".</p><p>Installation:<br></p><p>We use "tclap" as the command line tool which should be installed in the first step. Then you can compile BC.cu and main.cu to get the executable program.</p><p>You can run the executable program without params to see the usage document.<br></p> </div> </div>