基于贝叶斯最大熵分析的高维多目标优化方法

Many-objective optimization method based on Bayesian maximum entropy analysis

中文关键词  高维多目标优化  贝叶斯最大熵  参考点  空间随机场

英文关键词  many-objective optimization  Bayesian maximum entropy  reference point  random field

中文摘要

      实现收敛性和多样性之间的平衡是高维多目标优化面临的挑战之一.本文提出一种基于贝叶斯最大熵分 析的高维多目标优化方法,并通过与参考点策略和进化算法搜索操作相结合,实现了理想参考点和最差参考点的互 补协同.其主要思想是将高维目标空间中的解映射到空间随机场中,在硬数据和软数据参考序列的引导下,通过推 断和评估随机场中解的点位来应对不同Pareto前沿的优化问题.本文在基准DTLZ和MaF上,将所提出的方法与其他 4种高维多目标优化方法进行了比较,验证了所提出方法的可行性和有效性.

英文摘要

      Balancing convergence and diversity is a key challenge in many-objective optimization methods. A novel many-objective optimization framework based on the Bayesian maximum entropy is developed in the paper. It combines a reference point strategy with evolutionary algorithm search operations to achieve complementary and collaborative utiliza tion of both ideal and nadir reference points. The core idea is to map solutions from the high-dimensional objective space to a spatial random field. By leveraging both hard data and soft data reference sequences, it effectively tackles optimization problems with diverse Pareto frontiers by inferring and evaluating the positions of solutions in the random field. This paper compares the proposed method with four other many-objective optimization algorithms on benchmark DTLZ and MaF. The experimental results validate the feasibility and effectiveness of the proposed method.