Scored Pareto-MEC for multi-objective optimization

This paper proposes a new multi-objective optimization algorithm - scored Pareto mind evolutionary computation (SP-MEC), which introduces the theory of Pareto into mind evolutionary computation (MEC) for the multi-objective optimization. In our SP-MEC, the selection of individuals is based on their scores that include the Pareto dominance and density information among the individuals. The SP-MEC is compared with the VEGA, NSGA, SPEA, and Pareto-MEC on the basis of four different test problems: convexity, non-convexity, discreteness, as well as non-uniformity. Especially, both the Pareto-MEC and SPEA have shown promising performances in solving various optimization problems. On the test problems, SP-MEC outperforms all the four reference algorithms concerning three measures: the distance from trade-off front to Pareto-optimal front, the uniformity of solutions, and the spread of solutions. Impersonal termination criterion is used in SP-MEC and Pareto-MEC instead of the preset number of generations in other algorithms. SP-MEC has a higher computational efficiency than the VEGA, NSGA, and SPEA. Compared with another our algorithm, Pareto-MEC, the computational efficiency of SP-MEC is a little lower. However, the solution quality of SP-MEC is higher than that of the Pareto-MEC. Therefore, it can be concluded the SP-MEC is a powerful algorithm for multi-objective optimization.