I'm working to construct a learner than can recognize whether two vertices in a property graph (digraph, vertices and edges can have arbitrary keys/values) modelling a social network in fact represent the same person. I want to use the structure of the graph local to two vertices, compare those structures, and use the similarity as a feature.
I've had trouble finding related work. Light review of the literature leaves me stuck on graph methods - eg, Bayesian networks and graphs-as-learners- as opposed to learning graph-based domains. Are there any good resources I'm missing, or maybe suggested methods?