A cluster graph is a graph whose every connected component is a complete graph. Given a simple undirected graph $G$, a subset of vertices inducing a cluster graph is called an independent union of cliques (IUC), and the IUC polytope associated with $G$ is defined as the convex hull of the incidence vectors of all IUCs in the graph. The {\sc Maximum IUC} problem, which is to find a maximum-cardinality IUC in a graph, finds applications in network-based data analysis. In this paper, we derive several families of facet-defining valid inequalities for the IUC polytope. We also give a complete description of this polytope for some special classes of graphs. We establish computational complexity of the separation problem for most of the considered families of valid inequalities and explore the effectiveness of employing the corresponding cutting planes in an integer (linear) programming framework for the {\sc Maximum IUC} problem through computational experiments.
翻译:集束图是一张图,其中每个连接的组件都是完整的图形。如果有一个简单的非方向图形$G$,引起集束图的脊椎子集被称为独立的cliques联合体(IUC),与$G$相关的IUC聚变体被定义为图中所有IUCs发生矢量的圆柱体。在图表中找到最大心性IUC的问题,在基于网络的数据分析中找到应用。在本文中,我们得出了IUC聚点数个表面定义有效不平等的家族。我们还完整描述了某些特殊类别的图形。我们为大多数被认为具有有效不平等的家族确定了分离问题的计算复杂性,并探索了通过计算实验使用相应的剪切平面在 MSC 最大IUC}的整数(线性) 编程框架中使用相应的剪切面的有效性。