【作者】 李欣；

【导师】 罗琦；

【作者基本信息】 南京信息工程大学 ， 系统分析与集成， 2008， 硕士

【摘要】 简单遗传算法作为一种启发式搜索算法,寻优理论还不完善。因此,在应用中常出现收敛过慢、稳定性差及早熟现象等问题,而现有的一些自适应遗传算法容易产生局部最优解。因此,对自适应遗传算法的进一步研究和探讨是很必要的。针对简单遗传算法和现有的一些自适应遗传算法的缺陷,本文分析了种群“早熟”性能指标和计算量,并且判断种群当前适应度最大的那些个体是否重复或相互趋同,由此发展了一种新的种群“早熟”程度评价指标,结合自适应调整遗传算法的控制参数的思想,提出了一种改进的自适应遗传算法。作者希望本论文提出的新的自适应遗传算法,不仅能加快遗传进化速度,而且能增强遗传算法的全局收敛性能,从而得到满意的全局最优解。本文首先介绍了遗传算法的背景、发展历程和应用,国内外研究现状,说明了研究的背景、目的和预期结果;其次介绍了简单遗传算法和几种改进自适应遗传算法,分析了现有的一些自适应遗传算法存在的缺陷,为下一步工作奠定基础;最后本文提出了一种新的判定种群“早熟”程度的方法,对算法的交叉概率和变异概率进行改进,设计实现了本文提出的新算法。实验结果说明新算法具有计算稳定性高、收敛速度快等特点,是一种性能良好的改进的自适应遗传算法。更多还原

【Abstract】 Simple genetic algorithm is a heuristic searching algorithm. Its scheme of searching for the best result is not perfect. It has some shortages such as slow convergence, bad stability and premature phenomenon in application. Existing adaptive genetic algorithm has local optimization solution. Therefore, it is essential to implement a further research and discussion on the auto-adapted genetic algorithm.In order to solve the disadvantages of simple genetic algorithms and existing adaptive genetic algorithms, we began with analyzing the performance index and the computation load of the population "premature", and then developed a new degree evaluating indicator of new population "premature". It is expected that the present paper will propose a new self-adapted genetic algorithm to reach the satisfactory globally optimal solution, which can not only speed up the genetic evolution speed but also strengthen the corresponding global convergence performance.This article first introduced genetic algorithm’s background, the development process and the application, the domestic and foreign present research situation, the research background, the research goal and the anticipated result. Next this thesis explained the simple heredity algorithm and several kinds of improved auto-adapted genetic algorithm, moreover analyzed some existing flaw in the auto-adapted genetic algorithm, laying the foundation for the later work. Finally this article proposed a new kind of determination population "precocious" degree method. The involved idea is making the improvement to the algorithm overlapping probability and the variation probability. Furthermore, it is successful for the design to realize the new algorithm this article proposed. The experimental result showed that the new improved auto-adapted genetic algorithm is of better performance such as good stability, quick convergent speed and so on.更多还原