Parallel subgraph matching on a hierarchical interconnection network

  • Authors:

  • Stuart Campbell;Mohan Kumar;Horst Bunke

  • Affiliations:

  • Curtin Univ. of Technology, Perth, Australia;Curtin Univ. of Technology, Perth, Australia;Institute of Computer Science and Applied Mathematics, Bern, Switzerland

  • Venue:
  • Hardware implementation of intelligent systems
  • Year:
  • 2001

Quantified Score

Hi-index 0.00

Visualization

Abstract

The identification of subgraph isomorphisms is a well-known problem that occurs in many application areas. An important variant of the problem occurs when there are model graphs and a single input graph, and we wish to search for subgraph isomorphisms from any of the model graphs to the input graph. This chapter discusses the Parallel Netwprk (PN) algorithm; a parallel, deterministic algorithm for finding subgraph isomorphisms from a database of attributed model graphs to an attributed, directed input graph. The algorithm decomposes the model graphs and forms the resultant subgraphs into a number of search networks. Subgraphs common to any number of model graphs are represented only once. This approach allows rapid, parallel detection of matches of common subgraphs onto the input graph. In parallel, all mappings found for each model graph are searched to detect complete, consistent mappings, which define subgraph isomorphisms. When used on a hierarchical interconnection network, the algorithm allows local communication to be used to advantage, reducing communication overheads and improving performance.