解复杂优化问题的仿生算法的研究-运筹学与控制论专业论文.docxVIP

摘要、遗传算法和神经网络优化是新发展起来的仿生优化算法，它们分别借鉴了 摘要 、遗传算法和神经网络优化是新发展起来的仿生优化算法，它们分别借鉴了 生物的遗传进化机制和人脑的信息处理机制一本文提出了与试验设计相结合的 遗传算法，分别用于求解组合优化、多目标优化和非线性优化问题。另外，对 于种用神经网络求解组合优化问题的Ilybrid LT方法进行了改进，并对其进 行了收敛性分析。 首先本文介绍了遗传算法的基本框架，前人的理论成果，神经网络优化和 本文的研究工作，并埘全文各章节做了总体安排；然后简要介绍了正交设计与 均匀没计，并给出了这些试验设计方法与原有遗传算子结合的基本思路，它是 后面提出的新算法中构造遗传算子的理论基础；接着提出了求解优化问题的新 的遗传算法，主要包括求解组合优化问题、多目标优化问题和非线性优化问题 的遗传算法。此外，提出了两个概念：非劣解集的方差和熵，可定量描述不同 算法所得非劣解集的质量；在解非线性优化问题的遗传算法中，为保证算法的 有效性和种群的多样性，提出了一种新的选择方法，对于新算法的收敛性也做 了讨论。对Hybrid L，T算法进行了改进，并讨论了原算法没有讨论的收敛性问 题，得到了算法收敛的充分和必要条件。(最后对各算法进行了数值实验，结果 表明了各算法的有效性。、÷ 关键词：仿生算法遗传算法试验设计神经网络收敛‘娩 ABSTRACTGenetic ABSTRACT Genetic algorithms and neural network are new biologically-inspired optimization methods．These algorithms arc based o：c_-tain concepts of biological evolution and operation of the brain．In this paper some ne$v genetic algorithms based on experimental design are proposed．Besides，all improved Hybrid LT scheme iS presented and its convergence iS discussed． This paper starts with a basic structure of GAs and broad survey of cufTent techniques used in GAs．Chapter 2 includes a brief introduction of uniform design and orchogonal design，and discusses how to incorporate these experimental designs into evolutionary operators，which is the foundation of the resulting algorithms．Then several new genetic algorithms based on the experimental design are proposed．These algorithms includes algorithms for combinatorial optimization，multi-objective optimization and nonlinear optimization，and their convergences are studied． Moreover，two concepts：square—deviation and entropy of nondominated set are proposed．They Can quantitatively measure the quality of the nondominated set produced from difierent algorithms，In order to promote the eMciency of nonlinear GAs，a new selection is presented．This paper also gives an improved Hybrid L，r scheme．and jts convergence iS analyzed．At last．numerical experiments were executed for these algorithms and the results show the effectiveness of these algorithms．