摘要
Fast-Newman算法的复杂程度高,尤其是在计算模块度(Modularity)时,在边数较多的情况下,随着结点数提高,极大的影响着计算速度。为此,本文提出了一种基于Hadoop框架下的改进策略。该策略通过结点-边信息的划分,完成一定程度的分布化,在利用大量mappers的基础上,降低每次迭代时间,从而最终提升计算速度。通过对Zachary网络与随机ego-Facebook部分集的实验对比可以发现,算法加速比与并行序列数量有关。
To cut down the complexity of the fast-newman algorithm, especially the computation of 'modularity',which raises rapidly with the larger edges, a distributed fast-newman based on Hadoop framework has been proposed in this paper. It reduces the computing cost by degrading the number of pairs of edge and nodes to realize the computing parallel with matched count of mappers(computers). By recording the experiments of Zachary-net and the part of ego-Facebook, the relationship of speed-up ratio and numbers of mappers has been found.
出处
《科技广场》
2016年第11期9-12,共4页
Science Mosaic
