基于部分支配关系的双准则超多目标优化算法

A bi-criteria many-objective optimization algorithm based on partial dominance relations

近年来多目标优化问题受到研究者的广泛关注,也有多种解决方案应运而生。其中,多目标优化算法以其收敛速度快,种群保持收敛性和多样性好的特点而被广泛使用。但随着研究的展开,超多目标优化问题因其超高维的目标数导致现有多目标优化算法性能下降。本文充分利用基于帕累托(Pareto)和基于指标的算法优势,使用双准则方法以更好地处理超多目标优化问题。首先,考虑到基于Pareto的方法在高维目标空间的选择压力下降,通过部分支配关系提高算法对解的选择压力。其次,通过采用更加高效的互评价指标保证非Pareto部分的收敛速度和精度。最后,通过双准则进化实现两部分进化种群的信息交互和整合。通过参数实验和比较实验证明了所提方法的优越性。

Multi-objective optimization problems have drawn a lot of attention from researchers in recent years,and several solutions have been introduced.Among them,multi-objective optimization algorithms are widely used for their fast convergence speed and good population maintenance convergence and diversity.However,as the research develops,the performance of the existing multi-objective optimization algorithms degrades due to the ultra-high dimensional number of objectives.In this paper,we take advantage of the existing Pareto-based and metric-based algorithms and use a bi-criteria approach to better handle the many-objective optimization problems.First,considering the decreased selection pressure of the Pareto-based approach in the high-dimensional objective space,the selection pressure of the algorithm on the solution is improved through partial dominance relations.Then the convergence speed and accuracy of the non-Pareto part are guaranteed by using more efficient mutual evaluation metrics.Finally,the information interaction and integration of the two parts of the evolutionary population is achieved through bi-criterion evolution.In this paper,we will demonstrate the superiority of the proposed method through parametric and comparative experiments.