A framework for the study of preference incorporation in multiobjective evolutionary algorithms
Raluca Iordache, Serban Iordache, Florica Moldoveanu · 2014
We present a formal framework for the study of user preference incorporation into multiobjective evolutionary algorithms. This framework can accommodate virtually any preference model, including those that violate the independence of irrelevant alternatives. We also introduce the Preferanto notation, which permits the specification of a large variety of preference models. A number of properties and indicators are proposed for characterizing preference models. We report the results of a case study experiment assessing the impact of incorporating different preference models into an NSGA-II algorithm.