Distribution based representative sets for multi-objective integer programs

We study and exploit the characteristics of the nondominated sets of Multi-objective Integer Programs (MOIPs). We introduce a density measure and search for common properties of the distributions of nondominated points for different MOIPs. We design a procedure that categorizes the nondominated set into regions based on the densities of nondominated points. We develop an approach that generates representative sets of nondominated points using the estimated density information in different regions for general MOIPs. Experiments show that our approach is robust across different types of MOIPs.