arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2402.18015math.OC
收藏

通过分支定界法获得多目标优化问题的真 Pareto 最优解

Obtaining properly Pareto optimal solutions of multiobjective optimization problems via the branch and bound method

Weitian Wu, Xinmin Yang

AI总结:

本文提出一种基于分支定界的算法,利用凸多面体锥诱导的支配关系进行剪枝,为决策者提供 ε-真 Pareto 最优解,并证明了算法的全局收敛性,在基准问题和实际优化问题中验证了其有效性。

摘要:

在多目标优化中,大多数分支定界算法为决策者提供整个 Pareto 前沿,然后决策者最终可以从中选择一个解。然而,如果目标数量很大,候选解的数量也可能很大,决策者可能难以选择最感兴趣的解。正如我们在本文中所论述的,最感兴趣的解是那些权衡有界的解。这些解通常被称为真 Pareto 最优解。我们提出了一种基于分支定界的算法,为决策者提供所谓的 ε-真 Pareto 最优解。该算法的剪枝测试采用由凸多面体锥诱导的支配关系,而不是常用的 Pareto 支配关系。通过这种方式,所提出的算法将不包含 ε-真 Pareto 最优解的子盒从进一步探索中排除。我们建立了所提出算法的全局收敛性结果。最后,该算法被应用于基准问题以及两个实际优化问题。

英文摘要:

In multiobjective optimization, most branch and bound algorithms provide the decision maker with the whole Pareto front, and then decision maker could select a single solution finally. However, if the number of objectives is large, the number of candidate solutions may be also large, and it may be difficult for the decision maker to select the most interesting solution. As we argue in this paper, the most interesting solutions are the ones whose trade-offs are bounded. These solutions are usually known as the properly Pareto optimal solutions. We propose a branch-and-bound-based algorithm to provide the decision maker with so-called $ε$-properly Pareto optimal solutions. The discarding test of the algorithm adopts a dominance relation induced by a convex polyhedral cone instead of the common used Pareto dominance relation. In this way, the proposed algorithm excludes the subboxes which do not contain $ε$-properly Pareto optimal solution from further exploration. We establish the global convergence results of the proposed algorithm. Finally, the algorithm is applied to benchmark problems as well as to two real-world optimization problems.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑