Symmetry breaking in subgraph pattern matching

Zampelli, Stéphane [UCL] Deville, Yves [UCL] Dupont, Pierre [UCL]

Graph pattern matching, a central application in many fields, can be modelled as a CSP. This CSP approach can be competitive with dedicated algorithms. In this paper, we develop symmetry breaking techniques for subgraph matching in order to increase the number of tractable instances. Specific detection techniques are first developed for the classical variables symmetries and value symmetries. It is also shown how these symmetries can be broken when solving subgraph matching. We also show how conditional value symmetries can be automatically detected and handled in the search process. Then, the concept of local value symmetries is introduced; it is shown how these symmetries can be computed and exploited. Finally, experimental results show that symmetry breaking is an effective way to increase the number of tractable instances of the subgraph matching problem.