Iterative compaction: an improved approach to graph and circuit bisection

Abstract

Given a graph G=(V,E), graph bisection is the problem of finding a partition of the vertex set V in two equal-sized subsets V/sub 1/ and V/sub 2/ so that the number of edges between them is minimized. This problem has important applications in circuit partitioning, testing, VLSI design and other network-related problems that apply the divide-and-conquer strategy. The authors introduce a new heuristic approach, called iterative compaction (IC), which employees a node degree based matching and iterative graph compaction. This gives a significant improvement over the performance of known bisection algorithms in both time and quality of the results. >