Using Modified Determinantal Point Process Sampling to Update Population
Tengfei Li, Jinlong Li · 2018
Many-objective optimization problems (MaOPs) have posed a great challenge to the traditional Pareto-based multi-objective evolutionary algorithms (MOEAs) to balance convergence and diversity. To deal with this issue, we propose a Modified Determinantal Point Process Sampling (MDPPS) to select representatives from population. In order to evaluate the performance of our MDPPS, we plugged MDPPS into the Two Archive Algorithm in which the convergence archive and the diversity archive are updated by our MDPPS. Finally, the new two archive algorithm named TADPP is compared with five state-of-the-art algorithms on a variety of benchmark problems with different numbers of objectives. Experimental results demonstrate that TADPP is highly competitive on both convergence enhancement and diversity maintenance.