Efficient Mining of Frequent Subgraphs in the Presence of Isomorphism
ICDM '03 Proceedings of the Third IEEE International Conference on Data Mining
| Hi-index | 0.00 |
Many important algorithms in computational biology and related subjects rely on the ability to extract and to identify subgraphs of larger graphs; an example is to find common functional structures within Protein Interaction Networks. However, the increasing size of both the graphs to be searched and the target sub-graphs requires the use of large numbers of parallel conventional CPUs. This paper proposes an architecture to allow acceleration of sub-graph identification through reconfigurable hardware, using a canonical graph labelling algorithm. A practical implementation of the canonical labelling algorithm in the Virtex-4 reconfigurable architecture is presented, examining the scaling of resource usage and speed with changing algorithm parameters and input data-sets. The hardware labelling unit is over 100 times faster than a quad Opteron 2.2GHz for graphs with few vertex invariants, and at least 10 times faster for graphs that are easier to label.