Knowledge representation analysis of graph mining

• Published in: Annals of mathematics and artificial intelligence Vol. 86; no. 1-3; pp. 21 - 60
• Main Authors: van der Hallen, Matthias, Paramonov, Sergey, Janssens, Gerda, Denecker, Marc
• Format: Journal Article
• Language: English
• Published: Cham Springer International Publishing 01.07.2019; Springer; Springer Nature B.V
• Subjects: Adaptation; Analysis; Artificial Intelligence; Bioinformatics; Coding; Complex Systems; Computational linguistics; Computer Science; Data mining; Graphical representations; Knowledge representation; Language processing; Logic programs; Mathematics; Natural language interfaces; Pattern analysis; Social networks; Solvers; Specification and description languages; Specifications; Imperative declarative programming; Higher order; 68T30; Answer set programming; Knowledge representation; Graph mining
• DOI: 10.1007/s10472-019-09624-y
• ISSN: 1012-2443

This paper analyses the graph mining problem, and the frequent pattern mining task associated with it. In general, frequent pattern mining looks for a graph which occurs frequently within a network or, in the transactional setting, within a dataset of graphs. We discuss this task in the transactional setting, which is a problem of interest in many fields such as bioinformatics, chemoinformatics and social networks. We look at the graph mining problem from a Knowledge Representation point of view, hoping to learn something about support for higher-order logics in declarative languages and solvers. Graph mining is studied as a prototypical problem; it is easily expressible mathematically and exists in many variations. As such, it appears to be a prime candidate for a declarative approach; one would expect this allows for a clear, structured, statement of the problem combined with easy adaptation to changing requirements and variations. Current state-of-the-art KR languages such as IDP and ASP aspire to be practical solvers for such problems (Bruynooghe, Theory Practice Logic Program. (TPLP) 15 (6), 783–817 2015 ). Nevertheless, expressing the graph mining problem in these languages requires unexpectedly complicated and unintuitive encoding techniques. These techniques are in contrast to the ease with which one can transform the mathematical definition of graph mining to a higher-order logic specification, and distract from the problem essentials, complicating possible future adaptation. In this paper, we argue that efforts should be made towards supporting higher-order logic specifications in modern specification languages, without unintuitive and complicated encoding techniques. We argue that this not only makes representation clearer and more susceptible to future adaptation, but might also allow for faster, more competitive solver techniques to be implemented.