Title:An Efficient Pruning Process with Locality Aware Exploration and Dynamic Graph Editing for Subgraph Matching

Abstract:Subgraph matching is a NP-complete problem that extracts isomorphic embeddings of a query graph $q$ in a data graph $G$. In this paper, we present a framework with three components: Preprocessing, Reordering and Enumeration. While pruning is the core technique for almost all existing subgraph matching solvers, it mainly eliminates unnecessary enumeration over data graph without alternation of query graph. By formulating a problem: Assignment under Conditional Candidate Set(ACCS), which is proven to be equivalent to Subgraph matching problem, we propose Dynamic Graph Editing(DGE) that is for the first time designed to tailor the query graph to achieve pruning effect and performance acceleration. As a result, we proposed DGEE(Dynamic Graph Editing Enumeration), a novel enumeration algorithm combines Dynamic Graph Editing and Failing Set optimization. Our second contribution is proposing fGQL , an optimized version of GQL algorithm, that is utilized during the Preprocessing phase. Extensive experimental results show that the DGEE-based framework can outperform state-of-the-art subgraph matching algorithms.