Combining convergence and diversity in evolutionary multiobjective optimization
Evolutionary Computation
Comparison of Multiobjective Evolutionary Algorithms: Empirical Results
Evolutionary Computation
EMO'07 Proceedings of the 4th international conference on Evolutionary multi-criterion optimization
Evolutionary multi-objective optimization: a historical view of the field
IEEE Computational Intelligence Magazine
Multiobjective programming using uniform design and genetic algorithm
IEEE Transactions on Systems, Man, and Cybernetics, Part C: Applications and Reviews
An orthogonal genetic algorithm with quantization for globalnumerical optimization
IEEE Transactions on Evolutionary Computation
A fast and elitist multiobjective genetic algorithm: NSGA-II
IEEE Transactions on Evolutionary Computation
| Hi-index | 0.00 |
Multi-Objective Evolutionary Algorithms (MOEAs) are powerful and efficient tools to deal with multi-objective problems. In the framework of MOEAs, initialization is an important for the decision of the amount of space filling design information in first population. To fasten the convergence and heighten the hypervolume for MOEAs, in this paper, we adopt experimental methods to generate the first population including Orthogonal Design and Uniform Design. Compared with the traditional Random Design, the experimental methods can get well scattered solutions in feasible searching space and provide guiding information for the next offspring. In the experiment on bio-objective and triobjective problems by jMetal 2.0, we tested four state-of-art algorithms: NSGA-II, SPEA2, GDE3 and 2-MOEA. The results show that the orthogonal and uniform design outperforms the random design as it can significantly quicken the convergence and enhance the hypervolume. In addition, MOEAs with statistical initialization can obtain higher quality Pareto-optimal solutions in fewer numbers of fitness function evolutions.