东北大学学报:自然科学版 ›› 2017, Vol. 38 ›› Issue (5): 625-629.DOI: 10.12068/j.issn.1005-3026.2017.05.004

• 信息与控制 • 上一篇    下一篇

一种面向图集合的相似性搜索技术

庞俊, 谷峪, 于戈   

  1. (东北大学 计算机科学与工程学院, 辽宁 沈阳110819)
  • 收稿日期:2015-12-10 修回日期:2015-12-10 出版日期:2017-05-15 发布日期:2017-05-11
  • 通讯作者: 庞俊
  • 作者简介:庞俊(1984-),男,湖北咸宁人,东北大学博士研究生; 谷峪(1981-),男,辽宁鞍山人,东北大学教授,博士生导师; 于戈(1962-),男,辽宁大连人,东北大学教授,博士生导师.
  • 基金资助:

    国家重点基础研究发展计划项目(2012CB316201); 国家自然科学基金资助项目(61272179, 61472071).

A Similarity Search Technique for Graph Set

PANG Jun, GU Yu, YU Ge   

  1. School of Computer Science & Engineering, Northeastern University, Shenyang 110819, China.
  • Received:2015-12-10 Revised:2015-12-10 Online:2017-05-15 Published:2017-05-11
  • Contact: YU Ge
  • About author:-
  • Supported by:

    -

摘要:

目前图相似性的研究工作主要集中在子图的匹配,而没有充分关注图集合之间的匹配.针对这一问题,提出了一种基于过滤-求精框架的GSSS算法;提出了一种图集合距离定义,设计了Number,Size,Complete edge和Lower bound过滤器减小搜索空间,优化了图集合距离的计算;设计并优化了一种增量式的多层倒排索引,提高了查询效率,适应数据集的动态变化.真实数据集上的大量实验验证了GSSS算法的有效性和高效性.

关键词: 图集合, 相似性, 搜索, 索引, 过滤

Abstract:

Existing studies of graph similarity search mainly focus on the subgraph matching instead of the graph set matching. To tackle this issue, GSSS algorithm was proposed based on filtering-and-verify framework. A graph set distance was defined. In order to reduce the search space, Number filter, Size filter, Complete edge filter and Lower bound filter were proposed. Then, the computation of the graph set distance was optimized. An incremental multi-layer inverted index was designed to further improve the query efficiency. Extensive experiments on a real-world dataset show that GSSS algorithm is effective and efficient.

Key words: graph set, similarity, search, index, filter

中图分类号: