| 摘 要: 针对基于分解的多目标进化算法对帕累托前沿与权重向量分布的匹配度敏感、档案更新效率低以及解的多样性不足等问题,提出了一种潜力引导的权重自适应分解多目标进化算法(MOEA/D-PWA)。借鉴统计物理学整体势能概念构造存档选择目标函数,量化种群分布无序性以筛选高质量解；结合个体能量等级分类与交互机制,平衡存档多样性与收敛性；创造性地提出权重自适应更新策略,通过基于帕累托前沿估计删除冗余权重、结合存档种群稀疏性添加新权重的方式,精准指导动态权重调整。 |
| 关键词: 分解型多目标进化算法 帕累托前沿估计 势能指导 权重更新 |
|
中图分类号: TP391.4
文献标识码:
|
|
| Multi-Objective Evolutionary Algorithm for Adaptive Boosting Strategy |
|
WANG Zhiqiang
|
Jiangsu University of Science and Technology, Zhenjiang
|
| Abstract: Aiming at the problems that decomposition-based multi-objective evolutionary algorithms are sensitive to the matching degree between the Pareto front and weight vector distribution, suffer from low archive update efficiency, and lack solution diversity, a Potential-Guided Weight Adaptive Decomposition-Based Multi-Objective Evolutionary Algorithm (MOEA/D-PWA) is proposed. Drawing on the concept of global potential energy in statistical physics, an archive selection objective function is constructed to quantify the disorder of population distribution for screening high-quality solutions; by combining individual energy level classification and interaction mechanism, the balance between archive diversity and convergence is achieved; a weight adaptive update strategy is creatively proposed, which accurately guides dynamic weight adjustment by deleting redundant weights based on Pareto front estimation and adding new weights in combination with the sparsity of the archive population. |
| Keywords: MOEA/D Pareto frontier estimation potential energy guidance archive update |
