基于图编码的网络拓扑语义挖掘  

Network Topology Semantic Mining based on Graph Encoding

作　　者：杨扬[1] 张驰[1] YANG Yang;ZHANG Chi(Department of Electronic Engineering and Information Science,University of Science and Technology of China,Key Laboratory of Electromagnetic Space Information,Chinese Academy of Science,Hefei Anhui 230027,China)

机构地区：[1]中国科学技术大学电子工程与信息科学系中国科学院电磁空间信息重点实验室,安徽合肥230027

出　　处：《通信技术》2018年第11期2631-2638,共8页Communications Technology

基　　金：科技部国家重点研发计划"网络空间安全"重点专项(No.2017YFB0802202);国家自然科学基金项目(No.61702474)~~

摘　　要：对网络的拓扑语义理解,一直是网络结构信息挖掘的重要内容。当前的图分解和社区发现算法,对于网络中子图语义和子图间的重叠语义总是难以确定。基于最小描述长度的图编码方法,结合多样化的编码字典模板,能够统一地将网络中的拓扑结构语义进行挖掘和识别。实验分析表明,图编码方法能够优化网络结构理解,挖掘出网络的具体子图语义类型,有益于更深入地理解网络拓扑语义。The understanding on topology semantics of the network is always an important content for network structure information mining. The current graph partitioning and community detection algorithms are difficult to determine for subgraph semantics and overlapping semantics between subgraphs in the network. The graph encoding method based on the minimum description length combined with the diversified coding dictionary template can uniformly mine and identify the topology semantics in the network. The experimental analysis indicates that the graph coding method can optimize the understanding on network structure, mine semantic types of the specific subgraphs of the network, and deeply understand the topological semantics of the network.

关 键 词：最小描述长度 拓扑语义挖掘 图编码 图数据 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]