Nass: A New Approach to Graph Similarity Search

TL;DR

This paper introduces a novel graph similarity search method that improves candidate generation efficiency by leveraging GED computation results and a new GED algorithm, outperforming existing techniques on real datasets.

Contribution

The paper proposes a new approach that uses GED results to generate candidates and introduces a faster GED computation algorithm, addressing limitations of existing methods.

Findings

Significantly faster than state-of-the-art techniques

Reduces candidate set size as GED threshold increases

Effective on real-world datasets

Abstract

In this paper, we study the problem of graph similarity search with graph edit distance (GED) constraints. Due to the NP-hardness of GED computation, existing solutions to this problem adopt the filtering-and-verification framework with a main focus on the filtering phase to generate a small number of candidate graphs. However, they have a limitation that the number of candidates grows extremely rapidly as a GED threshold increases. To address the limitation, we propose a new approach that utilizes GED computation results in generating candidate graphs. The main idea is that whenever we identify a result graph of the query, we immediately regenerate candidate graphs using a subset of pre-computed graphs similar to the identified result graph. To speed up GED computation, we also develop a novel GED computation algorithm. The proposed algorithm reduces the search space for GED computation by utilizing a series of filtering techniques, which have been used to generate candidates in existing solutions. Experimental results on real datasets demonstrate the proposed approach significantly outperforms the state-of-the art techniques.