提高多目标进化算法分布性的动态调整机制

Dynamic diversity preservation strategy for multi-objective evolutionary algorithms

为提高多目标进化算法的分布性,提出一种基于极坐标的动态调整机制。在极坐标下,根据解集的拥挤程度,计算个体解的缩放系数。在进化过程中利用该缩放系数动态调整解集支配关系,适当提高分布性好的解在支配关系中的地位以改善解的分布。对测试函数的仿真试验结果表明,将该机制应用于经典算法能显著提高算法的分布性,同时保持良好的收敛性。

In order to improve diversity performance of multi objective evolutionary algorithms, a new dynamic diversity preservation strategy based on polar coordinates is proposed. Each solution is assigned a contract-expand coefficient which is related to its distribution in polar coordinates. This coefficient is used to adjust Pareto dominance in solutions set dynamically during evolution. Sparsely distributed solutions are evaluated in terms of Pareto dominance relation, which in turn improve the distribution of solutions set. Results show that the proposed strategy is able to improve conventional MOEAs on their diversity performance, at the same time, maintain convergence to Pareto optimal front on the same level.