信息与控制工程学院

Advanced Materials Research

中文关键字：结构模式识别；图嵌入；空间句法；K均值；拓扑；统计模式识别，英文关键字：structural pattern recognition; graph embedding; s

The main drawbacks of structural pattern recognition compared to statistical pattern recognition are the high computation complexity and fewer processing tools that are available in the domain. To bridge the gap between the structural and statistical pattern recognition, a new graph embedding method based on space syntax and improved K-means clustering is proposed. The present paper uses the space syntax theory to build quantitative description of the nodes’ topological features, and then combines the proposed topological features with non-topological features in other aspects of the domain to construct node feature set using an improved K-means clustering algorithm, and then maps the graph into vector space explicitly by a statistical approach. Thus SVM can be applied to achieve graph classification. The experimental results show that such an embedding method can achieve higher classification accuracy in different graph datasets.

李智杰

刘欣

zhengpuliang,lichanghua,lizhijie

Journal paper

卷：1044-1045

期：

页：1163-1168

no

2014-11-01