Sparsity-Aware Reachability Computation for Massive Graphs

Sung-Soo Kim, Young-Min Kang, Young-Kuk Kim

Abstract: We propose a novel sparsity-aware reachability computation framework so-called S-AORM, which provides fast performance for massive graphs via incremental fashion using sparse matrices. S-AORM is straightforward to compre-hend and outperforms existing methods in terms of compu-tational performance. Five synthetic networks generated from Barabási-Albert model and five real-world networks are used in comprehensive experiments. In terms of all-pairs shortest paths computation performance on the citation network, which is a directed and disconnected network, the proposed approach surpasses SNAP by up to 12.1 times and NetworkX by up to 80.2 times. The overall experimental results demonstrate that our approach provides significant performance improvement in the graph reachability computation.