没有合适的资源?快使用搜索试试~ 我知道了~
首页清华大学遗传算法毕业论文
资源详情
资源评论
资源推荐

清华大学毕业设计论文
摘 要
研究能在搜索过程中自动获取和积累有关搜索空间的知识,并自适应的控
制搜索过程,从而得到最优解或准最优解的通用搜索算法一直是令人瞩目的课
题。遗传算法就是这种特别有效的算法。
遗传算法(Genetic Algorithm——GA),是模拟达尔文的遗传选择和自然
淘汰的生物进化过程的计算模型,它是由美国 Michigan 大学的 J.Holland 教授
于 1975 年首先提出的。J.Holland 教授和它的研究小组围绕遗传算法进行研究的
宗旨有两个:抽取和解释自然系统的自适应过程以及设计具有自然系统机理的
人工系统。遗传算法的大致过程是这样的:将每个可能的解看作是群体中的一
个个体或染色体,并将每个个体编码成字符串的形式,根据预定的目标函数对
每个个体进行评价,即给出一个适应度值。开始时,总是随机的产生一些个体,
根据这些个体的适应度利用遗传算子——选择(Selection)、交 叉( Crossover)、
变异(Mutation)对它们重新组合,得到一群新的个体。这一群新的个体由于
继承了上一代的一些优良特性,明显优于上一代,以逐步向着更优解的方向进
化。遗传算法主要的特点在于:简单、通用、鲁棒性强。经过二十多年的发展,
遗传算法已经在旅行商问题、生产调度、函数优化、机器学习等领域得到成功
的应用。
传统的遗传算法虽然具有隐含的并行性,但其实实现方法在本质上仍然是
串行的。这种串行的遗传算法在解决一些实际问题时,由于需要较多的个体数
量和大量的计算,使得进化过程比较缓慢,难以达到实时的要求。因此并行遗
传算法(Parallel Genetic Alogrithm——PGA)就受到了较大的重视,并且已经
成为目前遗传算法研究的主要课题。遗传算法与并行计算机相结合,能把并行
机的高速性和遗传算法固有的并行性两者的长处彼此结合起来。
尽管遗传算法比其他传统搜索方法有更强的鲁棒性,但它更擅长全局搜索,
而局部搜索能力却不足。有研究证明,GA 可以用极快的速度达到最优解的 90%
左右,但要达到真正的最优解则要花费相当长的时间。实验表明,如果兼顾收
敛速度和解的质量两个指标,单纯的遗传算法未必能比其他搜索算法快很多。
目前较为活跃的研究领域是将遗传算法于传统的、基于问题知识的启发式搜索
i

清华大学毕业设计论文
技术相集成,即混合的遗传算法(Hybrid Genetic Algorithm——HGA)。典型 的
混合算法有 GA 与模拟退火算法的集成、混合算法与爬山法的集成等。
本文涉及了遗传算法的两个重要的应用领域:一个是利用混合并行遗传算
法解决连续优化问题——使用下山算法与传统的遗传算法相结合,对已经存在
的串行算法进行并行遗传算法的改造,并在机群系统并行计算机(8 个节点,每
个节点双 CPU)的 MPI 环境下对该算法进行测试,并对测试结果作适当的分析。
另一个是利用遗传算法解决组合优化问题——TSP(旅行商)问题,考察遗传算
法在求解 NP 问题中的性能。
关键词:遗传算法;混合遗传算法;并行遗传算法;下山算法;优化;TSP;
ii

清华大学毕业设计论文
Abstract
It has always been a compelling task to develop a universal algorithm
that can automatically gain and accumulate the knowledge of a search space
during the searching process and then find the best solution or approximately
the best solution. The Genetic Algorithm is such a kind of algorithm which
turn out to be very effective.
Genetic Algorithm is a part of evolutionary computing originally
invented by Dr J.Holland in the year 1975 and developed by him and his
students and colleges. As can be seen literally, Genetic Algorithm is inspired
by Darwins theory about evolution. GAs are inherently parallel optimization
procedures, which utilize a "survival of the fittest" approach. A basic GA
employs selection, crossover and mutation operators. The whole process can
be referred to as "reproduction". Algorithm is started with a set of solution
(represented by chromosomes) called sub-population. Solutions from one
sub-populations are taken and used to form a new sub-population by a hope
that the new population will be better, at least not worse than the old one.
Solutions are selected according to their fitness. The more suitable they are
the more chances they have to reproduce. The main characteristic of the GA is
that it is simple, universal and robust. After about 20 year’s development, GA
has been successfully used in a lot of fields such as the Travelling Salesman
Problem, Scheduling, Function Optimization, Machine Learning etc.
Traditional simple GAs have their own drawbacks. First, by mutation and
crossover, good sub-solution might be disrupted. Besides, maintaining
diversity seems to be prominent considerations. Because of these reasons, the
Parallel Genetic Algorithm (PGA) is taken into account. The parallel genetic
algorithm is an adaptation of the genetic algorithm to parallel computers.
Parallel search greatly overcome the problems in population diversity.
iii

清华大学毕业设计论文
Furthermore, with best local minimum broadcast to every node at certain
frequency, convergence time is remarkably reduced.
To enhance the convergence speed of a sub-population, Hybrid Genetic
Algorithm (such as genetic algorithm embedded with Downhill method) is
introduced. It has been very important in the application domain of function
optimization because it can greatly reduce the possibility of pre-mature. Two
different downhill method are implemented. Depth first downhill and width
first downhill.
This paper basically deals with two important applications of the Genetic
Algorithm. One is function optimization by Hybrid Parallel Genetic
Algorithm (HPGA) running under the MPI environment. Selection of genetic
factors is discussed and comparison with traditional GA is made to illustrate
the effectiveness of the algorithm. The other is Travelling Salesman Problem
by Parallel Genetic Algorithm.
Keywords: Genetic Algorithm; Hybrid Genetic Algorithm; Parallel Genetic
Algorithm; Downhill Method; Optimization; TSP;
iv

清华大学毕业设计论文
目 录
摘 要 I ········································································································
·····································································································
·······································································································
························································································
············································································
······················································································
································································································
···················································································
·························································································
·········································································
···············································································
··························································
·············································································
·······································································
···························································
····························································
······························································································
·······················································································
······························································································
·······················································································
·······················································································
······························································································
···························································································
··························································
ABSTRACT III
目 录 V
第一章 简单遗传算法 1
1.1 遗传算法的生物学基础 1
1.1.1 遗传和变异 1
1.1.2 进化 2
1.2 简单遗传算法简介 2
1.2.1 算法概要 2
1.2.2 遗传算法的手工模拟 6
1.3 简单遗传算法的特点 7
1.4 简单遗传算法的应用以及发展方向 10
1.4.1 遗传算法的应用 10
1.4.2 遗传算法的发展方向 11
第二章 遗传算法的基本原理和实现技术 13
2.1 模式定理(schemata theorem) 13
2.1.1 模式 13
2.1.2 模式定理 13
2.2 编码方法 16
2.2.1 编码问题 16
2.2.2 编码技术 16
2.3 群体设定 17
2.4 适应度函数 18
2.4.1 目标函数到适应度函数的映射 18
v
剩余61页未读,继续阅读













安全验证
文档复制为VIP权益,开通VIP直接复制

评论1