Learning Decision Variables in Many-Objective Optimization Problems

Traditional Multi-Objective Evolutionary Algorithms (MOEAs) have shown poor scalability in solving Many-Objective Optimization Problems (MaOPs). The use of machine learning techniques to enhance optimization algorithms applied to MaOPs has been drawing attention due to their ability to add domain knowledge during the search process. One method of this kind is inverse modeling, which uses machine learning models to enhance MOEAs differently, mapping the objective function values to the decision variables. The Decision Variable Learning (DVL) algorithm uses the inverse model in its concept and has shown good performance due to the ability to directly predict solutions closed to the Pareto-optimal front. The main goal of this work is to experimentally show the DVL as an optimization algorithm for MaOPs. Our results demonstrate that the DVL algorithm outperformed the NSGA-III, a well-known MOEA from the literature, in almost all scenarios with restriction on the number of objective functions with a high number of objectives.