Problem definitions for performance assessment of multi-objective optimization algorithms
V. L. Huang, AK Qin, Kalyanmoy Deb, Eckart Zitzler, Ponnuthurai Nagaratnam Suganthan, JJ Liang, Mike Preuß, Simon Huband · 2007
Optimizing multiple conflicting objectives results in more than one optimal solution (known as Pareto-optimal solutions). Although only one of these solutions will be adopted at the end, the recent trend in evolutionary and classical multi-objective optimization studies have focused on approximating the set of Pareto-optimal solutions. It is believed that such a set of solutions will collectively provide good insight to the different trade-off regions on the Pareto-optimal front, thereby aiding a better and more confident decision making at the end. However, the type of Pareto-optimal approximation being sought strongly depends on the decision maker; here, aspects such as convergence to the Pareto-optimal front and the maintenance of solution diversity are important. Thus, to assess the performance of such optimization algorithms, the preferences of the decision maker must be taken into account. Evolutionary Multi-objective Optimization (EMO) methodologies were suggested in the early 1990s for this task, and since then a number of performance assessment methods have been suggested. Most of the existing simulation studies that compare different EMO methodologies are based on a limited subset of performance measures. After more than 10 years of research and development into efficient EMO algorithms, the time is now ripe for the