Machine learning theory for graph pattern languages and its applications to graph mining

点击翻译按钮获取中文摘要

We proposed a series of techniques for extracting graph-structured patterns efficiently from graph-structured data, such as chemical compound data, HTML/XML data, network traffic data, and so on. In order to design expressive graph-structured patterns, we focused on tree-like properties (e. g., outerplanarity, tree-width) of target data. Our proposed algorithms run efficiently on real data, and discovered a number of interesting graph-structured patterns.