On the Solution of a Graph Partitioning Problem under Capacity Constraints

We study a variant of the graph partitioning problem where the weight of a cluster in the partition depends on the edges incident to its nodes. This problem was first proposed in the context of optical networks design. We recall complexity results and establish new inaproximability results. We then study several mixed integer quadratic programming formulations for the problem and different solutions techniques. We present experimental results comparing the various formulations and solution techniques.