Detecting changes in sequences of attributed graphs
Article in conference proceedings
In several application domains, one deals with data described by elements and their pair-wise relations. Accordingly, graph theory offers a sound framework to cast modeling and analysis tasks. However, due to a multitude of reasons, real-world systems change operating conditions over time, hence calling for time-dependent, graph-based representations of systems' state. Here, we deal with this problem by considering a methodology for detecting changes in sequences of graphs. The adopted methodology allows to process attributed graphs with variable order and topology, and for which a one-to-one vertex correspondence at different time steps is not given. In practice, changes are recognized by embedding each graph into a vector space, where conventional change detection procedures exist and can be easily applied. Theoretical results are presented in a companion paper. In this paper, we introduce the methodology and focus on expanding experimental evaluations on controlled yet relevant examples involving geometric graphs and Markov chains.
2017 IEEE Symposium Series on Computational Intelligence (SSCI) Proceedings
IEEE Symposium Series on Computational Intelligence
Honolulu, Hawaii, USA
27 Nov.-1 Dec. 2017
Change detection; Concept drift; Anomaly detection; Dynamic/Evolving graph; Attributed graph; Stationarity; Graph matching; Embedding