Modeling the maximum edge-weight k-plex partitioning problem

Given a sparse undirected graph G with weights on the edges, a k-plex partition of G is a partition of its set of nodes such that each component is a k-plex. A subset of nodes S is a k-plex if the degree of every node in the associated induced subgraph is at least |S|  k. The maximum edge-weight k-plex partitioning (Max-EkPP) problem is to find a k-plex partition with maximum total weight, where the partition’s weight is the sum of the weights on the edges in the solution. When k = 1, all components in the partition are cliques and the problem becomes the well-known maximum edge-weight clique partitioning (Max-ECP). However, and to our best knowledge, when k > 1, the problem has never been modeled. Actually, the literature on the k-plex addresses the search for a single component in an unweighted graph. We propose a polynomial size integer linear programming formulation for the Max-EkPP problem and consider the inclusion of additional topological constraints in the model. These constraints involve lower and upper limit capacity bounds in each component and upper bound constraints on the number of components in the final solution. All these characterizations preserve linearity and the initial polynomial size of the model. We also present computational tests in order to show the models’ performance under different parameters’ settings. These tests resort to benchmark and real-world graphs.