Declarative Approximate Graph Matching Using A Constraint Approach

Graph pattern matching is a central application in many fields. However, in many cases, the structure of the pattern can only be approximated and exact matching is then far too accurate. This work aims at proposing a CSP approach for approximate subgraph matching where the potential approximation is declaratively included in the pattern graph as optional nodes and forbidden edges. The model, covering both monomorphism and isomorphism problem, also allows additional properties, such as distance properties, between pairs of nodes in the pattern graph. Such properties can be either stated by the user, or automatically inferred by the system. The model is built through the definition of parametric morphism constraints, allowing an efficient implementation of propagators. An Oz/Mozart implementation has been developped. Experimental results show that our general framework is competitive with a specialized C++ Ullman (exact) matching algorithm, while also offering approximate matching.