Approximation techniques for hypergraph partitioning problems

Techniques for approximating a hypergraph by a weighted graph for use in node partitioning algorithms are described. The graphs use the same node set as the hypergraph and their edges are obtained by generating a series of cliques that correspond to subsets of the hyper edges. Edge weights are obtained by characterizing optimal solutions to mathematical programs that describe properties of feasible partitions of the hypergraph. These approximations are compared with other known approximations and yield promising results.