arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2510.02709cs.NE
收藏

一种使用指标驱动权重向量优化的多目标进化算法

A many-objective evolutionary algorithm using indicator-driven weight vector optimization

Qinghai Normal University(青海师范大学)

Xiaojing Han, Yuanxin Li

AI总结:

针对不规则帕累托前沿下固定权重向量导致的分布不均问题,提出基于MOEA/D框架、融合简化超体积指标与R2指标的自适应多目标进化算法,实现权重向量动态调整,实验证明其优于六种先进算法。

摘要:

对于规则帕累托前沿(PFs),例如平滑、连续且均匀分布的前沿,使用固定权重向量足以满足基于分解的多目标优化方法。然而,当遇到不规则PFs(包括退化、不连通、倒置等情况)时,固定权重向量通常会导致解集分布不均匀,甚至优化结果不佳。为解决此问题,本研究提出一种采用简化超体积指标的自适应多目标进化算法。该算法将指标评估技术与基于分解的方法相结合,促进多目标优化方法中权重向量的自适应动态调整。具体而言,基于MOEA/D框架,它使用简化超体积指标准确评估解的分布;同时,应用R2指标(作为超体积的近似)动态调节权重向量的更新频率。实验结果表明,与六种最先进算法相比,所提算法高效且有效。

英文摘要:

For regular Pareto Fronts (PFs), such as those that are smooth, continuous, and uniformly distributed, using fixed weight vectors is sufficient for multi-objective optimization approaches using decomposition. However, when encountering irregular PFs-including degenerate, disconnected, inverted, etc. Fixed weight vectors can often cause a non-uniform distribution of the sets or even poor optimization results. To address this issue, this study proposes an adaptive many-objective evolutionary algorithm with a simplified hypervolume indicator. It synthesizes indicator assessment techniques with decomposition-based methods to facilitate self-adaptive and dynamic adjustment of the weight vectors in many-objective optimization methods. Specifically, based on the MOEA/D framework, it uses a simplified hypervolume indicator to accurately assess solution distribution. Simultaneously, applying the R2 indicator (as an approximation of hypervolume) dynamically regulates the update frequency of the weight vectors. Experimental results demonstrate that the proposed algorithm is efficient and effective when compared with six state-of-the-art algorithms.

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。

↑