TopCom: Index for Shortest Distance Query in Directed Graph

Abstract

Finding shortest distance between two vertices in a graph is an important problem due to its numerous applications in diverse domains, including geo-spatial databases, social network analysis, and information retrieval. Classical algorithms (such as, Dijkstra) solve this problem in polynomial time, but these algorithms cannot provide real-time response for a large number of bursty queries on a large graph. So, indexing based solutions that pre-process the graph for efficiently answering (exactly or approximately) a large number of distance queries in real-time is becoming increasingly popular. Existing solutions have varying performance in terms of index size, index building time, query time, and accuracy. In this work, we propose TOPCOM, a novel indexing-based solution for exactly answering distance queries. Our experiments with two of the existing state-of-the-art methods (IS-Label and TreeMap) show the superiority of TOPCOM over these two methods considering scalability and query time. Besides, indexing of TOPCOM exploits the DAG (directed acyclic graph) structure in the graph, which makes it significantly faster than the existing methods if the SCCs (strongly connected component) of the input graph are relatively small.

Distance Indexing and Seed Clustering in Sequence Graphs

“…Many graph distance algorithms improve upon classical algorithms, such as Dijkstra's algorithm [4] and A* [7], by storing precomputed data in indexes. These methods index the identities of important edges [11,9] or distances between selected nodes [3,14,5,2] then use the indexed information to speed up distance calculations. Indexbased algorithms must make a tradeoff between the size of the index and the speed of the distance query.…”