目前,代理模型辅助的进化算法是提高复杂优化问题的计算效率的一种有效手段。其中,模型管理在代理辅助进化优化中起着至关重要的作用。提出了一种基于多目标加点规则的高斯过程模型辅助社会微粒群算法(Multi-objective infill criterion...目前,代理模型辅助的进化算法是提高复杂优化问题的计算效率的一种有效手段。其中,模型管理在代理辅助进化优化中起着至关重要的作用。提出了一种基于多目标加点规则的高斯过程模型辅助社会微粒群算法(Multi-objective infill criterion based Gaussian Process model assisted Social learning particle swarm optimization,MICGP-SLPSO)。将多目标的方法引入模型管理中,提出多目标加点规则,进而发展了一种新的基于代理模型的微粒群算法优化策略。选用高斯过程构造代理模型,采用微粒群算法对所构造的代理模型进行优化,根据已知信息,将期望改进准则(EI)及统计下限最小值准则LCB作为两个目标,用来确定哪些候选解进行实际计算。将本优化策略用于基准函数测试问题和阶梯悬臂梁设计优化实例,并与国内外现有研究成果进行比较,证明了MICGP-SLPSO在有限的适应值计算次数下拥有更好的寻优性能,尤其是在高维优化问题上拥有更显著的优势。展开更多
Equivalent simplification is an effective method for solving large-scale complex problems. In this paper, the authors simplify a classic project scheduling problem, which is the nonlinear continuous time-cost tradeoff...Equivalent simplification is an effective method for solving large-scale complex problems. In this paper, the authors simplify a classic project scheduling problem, which is the nonlinear continuous time-cost tradeoff problem(TCTP). Simplifying TCTP is a simple path problem in a critical path method(CPM) network. The authors transform TCTP into a simple activity float problem and design a complex polynomial algorithm for its solution. First, the authors discover relationships between activity floats and path lengths by studying activity floats from the perspective of path instead of time.Second, the authors perform simplification and improve the efficiency and accuracy of the solution by deleting redundant activities and narrowing the duration intervals of non-redundant activities. Finally,the authors compare our method with current methods. The relationships between activity floats and path lengths provide new approaches for other path and correlative project problems.展开更多
文摘目前,代理模型辅助的进化算法是提高复杂优化问题的计算效率的一种有效手段。其中,模型管理在代理辅助进化优化中起着至关重要的作用。提出了一种基于多目标加点规则的高斯过程模型辅助社会微粒群算法(Multi-objective infill criterion based Gaussian Process model assisted Social learning particle swarm optimization,MICGP-SLPSO)。将多目标的方法引入模型管理中,提出多目标加点规则,进而发展了一种新的基于代理模型的微粒群算法优化策略。选用高斯过程构造代理模型,采用微粒群算法对所构造的代理模型进行优化,根据已知信息,将期望改进准则(EI)及统计下限最小值准则LCB作为两个目标,用来确定哪些候选解进行实际计算。将本优化策略用于基准函数测试问题和阶梯悬臂梁设计优化实例,并与国内外现有研究成果进行比较,证明了MICGP-SLPSO在有限的适应值计算次数下拥有更好的寻优性能,尤其是在高维优化问题上拥有更显著的优势。
基金supported by the Science and Technology Foundation of Jiangxi Provincial Department of Education in China under Grant No.GJJ161114the Natural Science Foundation of China under Grant No.71271081the Soft Science Research Base of Water Security and Sustainable Development of Jiangxi Province in China
文摘Equivalent simplification is an effective method for solving large-scale complex problems. In this paper, the authors simplify a classic project scheduling problem, which is the nonlinear continuous time-cost tradeoff problem(TCTP). Simplifying TCTP is a simple path problem in a critical path method(CPM) network. The authors transform TCTP into a simple activity float problem and design a complex polynomial algorithm for its solution. First, the authors discover relationships between activity floats and path lengths by studying activity floats from the perspective of path instead of time.Second, the authors perform simplification and improve the efficiency and accuracy of the solution by deleting redundant activities and narrowing the duration intervals of non-redundant activities. Finally,the authors compare our method with current methods. The relationships between activity floats and path lengths provide new approaches for other path and correlative project problems.
