Rigorous computational methods for dimensionality reduction in multi-objective optimization

摘要

Multi-objective optimization (MOO) is an effective technique for studying optimal trade-off solutions that balance several criteria. The main limitation of MOO is that its computational burden grows in size with the number of objectives. With the goal to overcome this computational barrier, this work introduces a new algorithm for reducing the number of objectives in MOO. This approach is based on a mixed-integer linear programming (MILP) formulation presented previously by the authors that minimizes the error of omitting objectives in a MOO problem. We test the capabilities of this novel technique drawing a comparison with the original MILP and other two methods proposed in the literature.