基于consR的并行图匹配方法

Parallel Graph Matching Method Based on consR Algorithm

随着社交网络、生物网络规模的迅速扩大,能够快速、高效地实现对这些网络的匹配、查询等工作已经成为许多应用领域的迫切需求。给定两个网络图,图匹配的过程即为对图G1中的每个节点在图G2中找到唯一一个相对应的最为相似的节点,使得给定的两个图的匹配边的数量最多。文中基于大图匹配方法 consR,进行了两方面的优化:当图的节点数目较少时,优化了图G1、G2的相似性矩阵计算策略,从而使得图匹配的计算更加快捷;当图的节点数目较大时,针对匹配过程中最为耗时的步骤进行并行优化处理。实验结果表明,在与consR方法计算出的匹配结果保持一致的情况下,一定程度上缩短了图匹配计算时间。

With the rapid enlargement of the size of social network and biology network, how to implement the network matching and querying fast and efficiently has become an urgent need in many application research. Given two graphs Gl and G2 ,the task of graph mate- hing is to find,for each node in G1 ,the most similar node in G2 to maximize the number of matched edges. In this paper,based on the ef- fective large graph matching method consR,improve its efficiency from two aspects. When the number of nodes in the graph is relatively small, the computation cost for similarity matrix of G1 and G2 is reduced by a simplified yet effective computing strategy. When the number of nodes is large, adopt parallel strategy to speed up the matching procedure for large graphs. The experimental results show that the pro- posed method reduces the computation time of image matching at a certain extent while gets as good matching results as consR does.