遗传算法经典MATLAB代码

遗传算法详解（含MATLAB代码）Python遗传算法框架使用实例（一）使用Geatpy实现句子匹配在前面几篇文章中，我们已经介绍了高性能Python遗传和进化算法框架——Geatpy的使用。

本篇就一个案例进行展开讲述：pip install geatpy更新至Geatpy2的方法：pip install --upgrade --user geatpy查看版本号，在Python中执行：import geatpyprint(geatpy.__version__)我们都听过“无限猴子定理”，说的是有无限只猴子用无限的时间会产生特定的文章。

在无限猴子定理中，我们“假定”猴子们是没有像人类那样“智能”的，而且“假定”猴子不会自我学习。

因此，这些猴子需要“无限的时间"。

而在遗传算法中，由于采用的是启发式的进化搜索，因此不需要”无限的时间“就可以完成类似的工作。

当然，需要产生的文章篇幅越长，那么就需要越久的时间才能完成。

下面以产生"T om is a little boy, isn't he? Yes he is, he is a good and smart child and he is always ready to help others, all in all we all like him very much."的句子为例，讲述如何利用Geatpy实现句子的搜索。

之前的文章中我们已经讲述过如何使用Geatpy的进化算法框架实现遗传算法编程。

这里就直接用框架。

把自定义问题类和执行脚本编写在下面的"main.py”文件中：# -*- coding: utf-8 -*-import numpy as npimport geatpy as eaclass MyProblem(ea.Problem): # 继承Problem父类def __init__(self):name = 'MyProblem' # 初始化name（函数名称，可以随意设置） # 定义需要匹配的句子strs = 'Tom is a little boy, isn't he? Yes he is, he is a good and smart child and he is always ready to help others, all in all we all like him very much.'self.words = []for c in strs:self.words.append(ord(c)) # 把字符串转成ASCII码M = 1 # 初始化M（目标维数）maxormins = [1] # 初始化maxormins（目标最小最大化标记列表，1：最小化该目标；-1：最大化该目标）Dim = len(self.words) # 初始化Dim（决策变量维数）varTypes = [1] * Dim # 初始化varTypes（决策变量的类型，元素为0表示对应的变量是连续的；1表示是离散的）lb = [32] * Dim # 决策变量下界ub = [122] * Dim # 决策变量上界lbin = [1] * Dim # 决策变量下边界ubin = [1] * Dim # 决策变量上边界# 调用父类构造方法完成实例化ea.Problem.__init__(self, name, M, maxormins, Dim, varTypes, lb, ub, lbin, ubin)def aimFunc(self, pop): # 目标函数Vars = pop.Phen # 得到决策变量矩阵diff = np.sum((Vars - self.words)**2, 1)pop.ObjV = np.array([diff]).T # 把求得的目标函数值赋值给种群pop的ObjV执行脚本if __name__ == "__main__":"""================================实例化问题对象============================="""problem = MyProblem() # 生成问题对象"""==================================种群设置================================"""Encoding = 'RI' # 编码方式NIND = 50 # 种群规模Field = ea.crtfld(Encoding, problem.varTypes, problem.ranges,problem.borders) # 创建区域描述器population = ea.Population(Encoding, Field, NIND) # 实例化种群对象（此时种群还没被初始化，仅仅是完成种群对象的实例化）"""================================算法参数设置=============================="""myAlgorithm = ea.soea_DE_rand_1_L_templet(problem, population) # 实例化一个算法模板对象myAlgorithm.MAXGEN = 2000 # 最大进化代数"""===========================调用算法模板进行种群进化========================="""[population, obj_trace, var_trace] = myAlgorithm.run() # 执行算法模板population.save() # 把最后一代种群的信息保存到文件中# 输出结果best_gen = np.argmin(obj_trace[:, 1]) # 记录最优种群是在哪一代best_ObjV = obj_trace[best_gen, 1]print('最优的目标函数值为：%s'%(best_ObjV))print('有效进化代数：%s'%(obj_trace.shape[0]))print('最优的一代是第 %s 代'%(best_gen + 1))print('评价次数：%s'%(myAlgorithm.evalsNum))print('时间已过 %s 秒'%(myAlgorithm.passTime))for num in var_trace[best_gen, :]:print(chr(int(num)), end = '')上述代码中首先定义了一个问题类MyProblem，然后调用Geatpy内置的soea_DE_rand_1_L_templet算法模板，它实现的是差分进化算法DE-rand-1-L，详见源码：运行结果如下：种群信息导出完毕。

function youhuafunD=code;N=50; % Tunablemaxgen=50; % Tunablecrossrate=0.5; %Tunablemuterate=0.08; %Tunablegeneration=1;num = length(D);fatherrand=randint(num,N,3);score = zeros(maxgen,N);while generation<=maxgenind=randperm(N-2)+2; % 随机配对交叉A=fatherrand(:,ind(1:(N-2)/2));B=fatherrand(:,ind((N-2)/2+1:end));% 多点交叉rnd=rand(num,(N-2)/2);ind=rnd tmp=A(ind);A(ind)=B(ind);B(ind)=tmp;% % 两点交叉% for kk=1:(N-2)/2% rndtmp=randint(1,1,num)+1;% tmp=A(1:rndtmp,kk);% A(1:rndtmp,kk)=B(1:rndtmp,kk);% B(1:rndtmp,kk)=tmp;% endfatherrand=[fatherrand(:,1:2),A,B];% 变异rnd=rand(num,N);ind=rnd [m,n]=size(ind);tmp=randint(m,n,2)+1;tmp(:,1:2)=0;fatherrand=tmp+fatherrand;fatherrand=mod(fatherrand,3);% fatherrand(ind)=tmp;%评价、选择scoreN=scorefun(fatherrand,D);% 求得N个个体的评价函数score(generation,:)=scoreN;[scoreSort,scoreind]=sort(scoreN);sumscore=cumsum(scoreSort);sumscore=sumscore./sumscore(end);childind(1:2)=scoreind(end-1:end);for k=3:N tmprnd=rand;tmpind=tmprnd difind=[0,diff(tmpind)];if ~any(difind)difind(1)=1;endchildind(k)=scoreind(logical(difind));endfatherrand=fatherrand(:,childind);generation=generation+1;end% scoremaxV=max(score,[],2);minV=11*300-maxV;plot(minV,'*');title('各代的目标函数值');F4=D(:,4);FF4=F4-fatherrand(:,1);FF4=max(FF4,1);D(:,5)=FF4;save DData Dfunction D=codeload youhua.mat% properties F2 and F3F1=A(:,1);F2=A(:,2);F3=A(:,3);if (max(F2)>1450)||(min(F2)<=900)error('DATA property F2 exceed it''s range (900,1450]') end% get group property F1 of data, according to F2 value F4=zeros(size(F1));for ite=11:-1:1index=find(F2<=900+ite*50);F4(index)=ite;endD=[F1,F2,F3,F4];function ScoreN=scorefun(fatherrand,D)F3=D(:,3);F4=D(:,4);N=size(fatherrand,2);FF4=F4*ones(1,N);FF4rnd=FF4-fatherrand;FF4rnd=max(FF4rnd,1);ScoreN=ones(1,N)*300*11;% 这里有待优化for k=1:NFF4k=FF4rnd(:,k);for ite=1:11F0index=find(FF4k==ite);if ~isempty(F0index)tmpMat=F3(F0index);tmpSco=sum(tmpMat);ScoreBin(ite)=mod(tmpSco,300);endendScorek(k)=sum(ScoreBin);endScoreN=ScoreN-Scorek;遗传算法实例:% 下面举例说明遗传算法%% 求下列函数的最大值%% f(x)=10*sin(5x)+7*cos(4x) x∈[0,10] %% 将x 的值用一个10位的二值形式表示为二值问题，一个10位的二值数提供的分辨率是每为(10-0)/(2^10-1)≈0.01 。

nsga-ⅲ算法matlab代码及注释一、NSGA-Ⅲ算法简介NSGA-III算法是多目标优化领域的一种经典算法，它是基于非支配排序的遗传算法。

该算法通过模拟自然选择的过程，不断改进种裙中的个体，以寻找Pareto前沿上的最优解。

NSGA-III算法在解决多目标优化问题方面表现出色，广泛应用于工程、经济和管理等领域。

二、代码实现下面是NSGA-III算法的Matlab代码示例，包含了代码的注释和解释。

```matlab初始化参数pop_size = 100; 种裙大小max_gen = 100; 最大迭代次数p_cross = 0.8; 交叉概率p_mut = 0.1; 变异概率n_obj = 2; 目标函数数量初始化种裙pop = initialization(pop_size);进化过程for gen = 1:max_gen非支配排序和拥挤度距离计算[fronts, cd] = non_dominated_sort(pop);种裙选择offspring = selection(pop, fronts, cd, pop_size);交叉和变异offspring = crossover(offspring, p_cross);offspring = mutation(offspring, p_mut);合并父代和子代种裙pop = merge_pop(pop, offspring, pop_size);end结果分析pareto_front = get_pareto_front(pop);plot_pareto_front(pareto_front);```三、代码解释1. 初始化参数：设置种裙大小、最大迭代次数、交叉概率、变异概率和目标函数数量等参数。

2. 初始化种裙：调用初始化函数，生成初始的种裙个体。

3. 进化过程：在每一代中，进行非支配排序和拥挤度距离计算，然后进行种裙选择、交叉和变异操作，最后合并父代和子代种裙。

遗传算法MATLAB完整代码（不用工具箱）遗传算法解决简单问题%主程序：用遗传算法求解y=200*exp(-0.05*x).*sin(x)在区间[-2,2]上的最大值clc;clear all;close all;global BitLengthglobal boundsbeginglobal boundsendbounds=[-2,2];precision=0.0001;boundsbegin=bounds(:,1);boundsend=bounds(:,2);%计算如果满足求解精度至少需要多长的染色体BitLength=ceil(log2((boundsend-boundsbegin)'./precision));popsize=50; %初始种群大小Generationmax=12; %最大代数pcrossover=0.90; %交配概率pmutation=0.09; %变异概率%产生初始种群population=round(rand(popsize,BitLength));%计算适应度，返回适应度Fitvalue和累计概率cumsump[Fitvalue,cumsump]=fitnessfun(population);Generation=1;while Generation<generationmax+1< p="">for j=1:2:popsize%选择操作seln=selection(population,cumsump);%交叉操作scro=crossover(population,seln,pcrossover);scnew(j,:)=scro(1,:);scnew(j+1,:)=scro(2,:);%变异操作smnew(j,:)=mutation(scnew(j,:),pmutation);smnew(j+1,:)=mutation(scnew(j+1,:),pmutation);endpopulation=scnew; %产生了新的种群%计算新种群的适应度[Fitvalue,cumsump]=fitnessfun(population);%记录当前代最好的适应度和平均适应度[fmax,nmax]=max(Fitvalue);fmean=mean(Fitvalue);ymax(Generation)=fmax;ymean(Generation)=fmean;%记录当前代的最佳染色体个体x=transform2to10(population(nmax,:));%自变量取值范围是[-2,2],需要把经过遗传运算的最佳染色体整合到[-2,2]区间xx=boundsbegin+x*(boundsend-boundsbegin)/(power((boundsend),BitLength)-1);xmax(Generation)=xx;Generation=Generation+1;endGeneration=Generation-1;Bestpopulation=xx;Besttargetfunvalue=targetfun(xx);%绘制经过遗传运算后的适应度曲线。

遗传算法matlab程序代码遗传算法是一种优化算法，用于在给定的搜索空间中寻找最优解。

在Matlab中，可以通过以下代码编写一个基本的遗传算法：% 初始种群大小Npop = 100;% 搜索空间维度ndim = 2;% 最大迭代次数imax = 100;% 初始化种群pop = rand(Npop, ndim);% 最小化目标函数fun = @(x) sum(x.^2);for i = 1:imax% 计算适应度函数fit = 1./fun(pop);% 选择操作[fitSort, fitIndex] = sort(fit, 'descend');pop = pop(fitIndex(1:Npop), :);% 染色体交叉操作popNew = zeros(Npop, ndim);for j = 1:Npopparent1Index = randi([1, Npop]);parent2Index = randi([1, Npop]);parent1 = pop(parent1Index, :);parent2 = pop(parent2Index, :);crossIndex = randi([1, ndim-1]);popNew(j,:) = [parent1(1:crossIndex),parent2(crossIndex+1:end)];end% 染色体突变操作for j = 1:NpopmutIndex = randi([1, ndim]);mutScale = randn();popNew(j, mutIndex) = popNew(j, mutIndex) + mutScale;end% 更新种群pop = [pop; popNew];end% 返回最优解[resultFit, resultIndex] = max(fit);result = pop(resultIndex, :);以上代码实现了一个简单的遗传算法，用于最小化目标函数x1^2 + x2^2。

方案一的程序编码函数主文件：function[Xp,LC1,LC2,LC3]=CLBGA8(M,Pm) %%%陈璐斌编程，解决VRP问题（带时间窗）%%输入参数%M遗传进化迭代次数%Pm变异概率%%输出参数%Xp最优个体%LC1目标收敛曲线%LC2平均适应度收敛曲线%LC3最优适应度收敛曲线%%%变量初始化Xp=zeros(1,5);LC1=zeros(1,M);LC2=zeros(1,M);LC3=zeros(1,M);Best=inf;%%编码方式-第一步：产生初始种群N=10;%N 种群规模farm=cell(1,N);%存储种群的细胞结构k=1;while (N-k>=0)G=randperm(5);%产生5个客户的全排列farm{k}=G;k=k+1;end%%%进化迭代计数器counter=1;while counter<=M%%第二步：交叉%交叉采用双亲双子单点交叉N=10;%种群规模newfarm=cell(1,2*N-4);%存储子代的细胞结构Ser=randperm(N);%两两随机配对表生成for i=1:(N-2)%避免交叉概率为1 A=farm{Ser(i)};B=farm{Ser(i+1)};%取出父代P0=unidrnd(5);%随机选择交叉点aa=zeros(1,5);bb=zeros(1,5);A_=A;B_=B;for ii=1:5-P0aa(ii)=B(P0+ii);endfor ii=1:5-P0for iiii=1:5if(B(P0+ii)==A_(iiii))A_(iiii)=0;endendendfor iii=6-P0:5for iiii=1:5if(A_(iiii)~=0)aa(iii)=A_(iiii);A_(iiii)=0;breakendendendfor ii=1:5-P0bb(ii)=A(P0+ii);endfor ii=1:5-P0for iiii=1:5if(A(P0+ii)==B_(iiii))B_(iiii)=0;endendendfor iii=6-P0:5for iiii=1:5if(B_(iiii)~=0)bb(iii)=B_(iiii);B_(iiii)=0;breakendendend%产生子代newfarm{2*i-1}=aa;newfarm{2*i}=bb;endFARM=[farm,newfarm];%新旧种群合并%%第三步：选择复制%%计算当前种群适应度并存储N=10;SYZ=zeros(1,3*N-4);syz=zeros(1,3*N-4);for i=1:(3*N-4)x=FARM{i};SYZ(i)=clb8(x);end%%选择复制，较优的N个个体复制到下一代k=1;while k<=(3*N-4)maxSYZ=max(SYZ);posSYZ=find(SYZ==maxSYZ);POS=posSYZ(1);k=k+1;farm{k}=FARM{POS};syz(k)=SYZ(POS);SYZ(POS)=0;end%记录和更新，更新最优个体，记录收敛曲线数据maxsyz=max(syz);meansyz=mean(syz);pos=find(syz==maxsyz);LC2(counter+1)=meansyz;if maxsyzBest=maxsyz;Xp=farm{pos(1)};endLC3(counter+1)=Best;d=[0,6.4,3.2,3.9,3.7,2;6.4,0,2.9,2.1,4.5,4.1;3.2,2.9,0,1.5,3.3,1.2;3.9,2.1,1.5,0,3.6,2.6;3.7,4.5,3.3,3.6 ,0,3.8;...2.0,4.1,1.2,2.6,3.8,0;];%距离矩阵t=[0,0.16,0.08,0.1,0.09,0.05;0.16,0,0.07,0.05,0.11,0.1;0.08,0.07,0,0.04,0.08,0.03;...0.1,0.05,0.04,0,0.09,0.07;0.09,0.11,0.08,0.09,0,0.10;0.05,0.1,0.03,0.07,0.1,0;];%行驶时间矩阵w=[0.15,0.2,0.18,0.25,0.22];%服务时间矩阵%%时间窗向量early=[0.15,0.3,0.7,0.4,0.7];xx=x;%取出染色体j=1;%分工点初始化%%取距离向量d1,d2d1=zeros(1,6);d1(1)=d(1,xx(1)+1);for i=1:4d1(i+1)=d(xx(i)+1,xx(i+1)+1);endd1(6)=d(xx(5)+1,1);%%时间窗计算T=t(1,xx(1)+1);pun1=0;if T<early(xx(1))pun1=early(xx(1))-T;T=early(xx(1));endT=T+w(xx(1));for i=2:5T=T+t(xx(i-1)+1,xx(i)+1);if T<early(xx(i))pun1=pun1+early(xx(i))-T;T=early(xx(i));endT=T+w(xx(5));endF=sum(10.*d1)+sum(10.*d2)+20*pun1; LC1(counter+1)=F;%%第四步：变异N=10;for i=1:Nif Pm>randAA=farm{i};POS1=unidrnd(5);POS2=unidrnd(5);temp=AA(POS1);AA(POS1)=AA(POS2);AA(POS2)=temp;farm{i}=AA;endendcounter=counter+1;end%%第五步：绘制收敛曲线图figure(2);plot(LC1);xlabel('迭代次数');ylabel('目标的值');title('目标的收敛曲线');figure(3);plot(LC2);xlabel('迭代次数');ylabel('适应度函数的平均值');title('平均适应度函数的收敛曲线');plot(LC3);xlabel('迭代次数');ylabel('适应度函数的最优值');title('最优适应度函数的收敛曲线');适应度文件：%%计算载重量和时间窗%%适应度函数计算function Fitness=clb8(x)d=[0,6.4,3.2,3.9,3.7,2;6.4,0,2.9,2.1,4.5,4.1;3.2,2.9,0,1.5,3.3,1.2;3.9,2.1,1.5,0,3.6,2.6;3.7,4.5,3.3,3.6 ,0,3.8;...2.0,4.1,1.2,2.6,3.8,0;];%距离矩阵t=[0,0.16,0.08,0.1,0.09,0.05;0.16,0,0.07,0.05,0.11,0.1;0.08,0.07,0,0.04,0.08,0.03;...0.1,0.05,0.04,0,0.09,0.07;0.09,0.11,0.08,0.09,0,0.10;0.05,0.1,0.03,0.07,0.1,0;];%行驶时间矩阵w=[0.15,0.2,0.18,0.25,0.22];%服务时间矩阵%%时间窗向量early=[0.15,0.3,0.7,0.4,0.7];xx=x;%取出染色体j=1;%分工点初始化%%取距离向量d1,d2d1=zeros(1,6);d1(1)=d(1,xx(1)+1);for i=1:4d1(i+1)=d(xx(i)+1,xx(i+1)+1);endd1(6)=d(xx(5)+1,1);%%时间窗计算T=t(1,xx(1)+1);pun1=0;if T<early(xx(1))pun1=early(xx(1))-T;T=early(xx(1));endT=T+w(xx(1));T=T+t(xx(i-1)+1,xx(i)+1);if T<early(xx(i))pun1=pun1+early(xx(i))-T;T=early(xx(i));endT=T+w(xx(5));endF=sum(10.*d1)+sum(10.*d2)+20*pun1;Fitness=1/F;计算时间文件：function[T]=TOTALT(Xp1)Xp=Xp1;t=[0,0.16,0.08,0.1,0.09,0.05;0.16,0,0.07,0.05,0.11,0.1;0.08,0.07,0,0.04,0.08,0.03;...0.1,0.05,0.04,0,0.09,0.07;0.09,0.11,0.08,0.09,0,0.10;0.05,0.1,0.03,0.07,0.1,0;];%行驶时间矩阵w=[0.15,0.2,0.18,0.25,0.22];%服务时间矩阵%%时间窗向量early=[0.15,0.3,0.7,0.4,0.7];T=t(1,Xp(1)+1);if T<early(Xp(1))T=early(Xp(1));endT=T+w(Xp(1));for i=2:5T=T+t(Xp(i-1)+1,Xp(i)+1);if T<early(Xp(i))T=early(Xp(1));endT=T+w(Xp(i));endT=T+t(1,Xp(5)+1);方案二的程序编码主函数文件：function[Xp,LC1,LC2,LC3]=CLBGA9(M,Pm)%%%陈璐斌编程，解决VRP问题（带时间窗）%%输入参数%M遗传进化迭代次数%Pm变异概率%%输出参数%Xp最优个体%LC1子目标2收敛曲线%LC2平均适应度收敛曲线%LC3最优适应度收敛曲线%%%变量初始化Xp=zeros(1,6);LC1=zeros(1,M);LC2=zeros(1,M);LC3=zeros(1,M);Best=inf;%%编码方式-第一步：产生初始种群N=10;%N 种群规模%Q=[2.4,3.3,2.1,2.7,2.3,1.6,2.0,1.2,3.6,1.9];%需求矩阵farm=cell(1,N);%存储种群的细胞结构k=1;while (N-k>=0)G=randperm(6);%产生6个客户的全排列farm{k}=G;k=k+1;end%%%进化迭代计数器counter=1;while counter<=M%%第二步：交叉%交叉采用双亲双子单点交叉N=10;%种群规模newfarm=cell(1,2*N-4);%存储子代的细胞结构Ser=randperm(N);%两两随机配对表生成for i=1:(N-2)%避免交叉概率为1A=farm{Ser(i)};B=farm{Ser(i+1)};%取出父代P0=unidrnd(6);%随机选择交叉点aa=zeros(1,6);bb=zeros(1,6);A_=A;B_=B;for ii=1:6-P0aa(ii)=B(P0+ii);endfor ii=1:6-P0for iiii=1:6if(B(P0+ii)==A_(iiii))A_(iiii)=0;endendendfor iii=7-P0:6for iiii=1:6if(A_(iiii)~=0)aa(iii)=A_(iiii);A_(iiii)=0;breakendendendfor ii=1:6-P0bb(ii)=A(P0+ii);endfor ii=1:6-P0for iiii=1:6if(A(P0+ii)==B_(iiii))B_(iiii)=0;endendendfor iii=7-P0:6for iiii=1:6if(B_(iiii)~=0)bb(iii)=B_(iiii);B_(iiii)=0;breakendendend%产生子代newfarm{2*i-1}=aa;newfarm{2*i}=bb;endFARM=[farm,newfarm];%新旧种群合并%%第三步：选择复制%%计算当前种群适应度并存储N=10;SYZ=zeros(1,3*N-4);syz=zeros(1,3*N-4);for i=1:(3*N-4)x=FARM{i};SYZ(i)=clb9(x);end%%选择复制，较优的N个个体复制到下一代k=1;while k<=(3*N-4)maxSYZ=max(SYZ);posSYZ=find(SYZ==maxSYZ);POS=posSYZ(1);k=k+1;farm{k}=FARM{POS};syz(k)=SYZ(POS);SYZ(POS)=0;end%记录和更新，更新最优个体，记录收敛曲线数据maxsyz=max(syz);meansyz=mean(syz);pos=find(syz==maxsyz);LC2(counter+1)=meansyz;if maxsyzBest=maxsyz;Xp=farm{pos(1)};endLC3(counter+1)=Best;d=[0,6.4,3.2,3.9,3.7,35,2;6.4,0,2.9,2.1,4.5,32.5,4.1;3.2,2.9,0,1.5,3.3,35.7,1.2;3.9,2.1,1.5,0,3.6,34.5,2.6;...3.7,4.5,3.3,3.6,0,37,3.8;35,32.5,35.7,34.5,37,0,38.5;2,4.1,1.2,2.6,3.8,38.5,0];%距离矩阵t=[0,0.16,0.08,0.1,0.1,0.88,0.05;0.16,0,0.07,0.05,0.11,0.81,0.1;0.08,0.07,0,0.04,0.08,0.9,0.03;...0.1,0.05,0.04,0,0.09,0.86,0.07;0.1,0.11,0.08,0.09,0,0.92,0.1;0.88,0.81,0.9,0.86,0.92,0,0.96;...0.05,0.1,0.03,0.07,0.1,0.96,0;];%行驶时间矩阵w=[0.15,0.2,0.18,0.25,0.2,0.22];%服务时间矩阵%%时间窗向量early=[0.15,0.3,0.7,0.4,0.7,0.6];xx=x;%取出染色体j=1;%分工点初始化%%取距离向量d1,d2d1=zeros(1,7);d1(1)=d(1,xx(1)+1);for i=1:5d1(i+1)=d(xx(i)+1,xx(i+1)+1);endd1(7)=d(xx(6)+1,1);%%时间窗计算T=t(1,xx(1)+1);pun1=0;if T<early(xx(1))pun1=early(xx(1))-T;T=early(xx(1));endT=T+w(xx(1));for i=2:6T=T+t(xx(i-1)+1,xx(i)+1);if T<early(xx(i))pun1=pun1+early(xx(i))-T;T=early(xx(i));endT=T+w(xx(6));endF=sum(10.*d1) +20*pun1;LC1(counter+1)=F;%%第四步：变异N=10;for i=1:Nif Pm>randAA=farm{i};POS1=unidrnd(6);POS2=unidrnd(6);temp=AA(POS1);AA(POS1)=AA(POS2);AA(POS2)=temp;farm{i}=AA;endendcounter=counter+1;end%%第五步：绘制收敛曲线图figure(2);plot(LC1);xlabel('迭代次数');ylabel('目标的值');title('目标的收敛曲线');figure(3);plot(LC2);xlabel('迭代次数');ylabel('适应度函数的平均值');title('平均适应度函数的收敛曲线');figure(4);plot(LC3);xlabel('迭代次数');ylabel('适应度函数的最优值');title('最优适应度函数的收敛曲线');适应度文件：%%计算载重量和时间窗%%适应度函数计算function Fitness=clb9(x)d=[0,6.4,3.2,3.9,3.7,35,2;6.4,0,2.9,2.1,4.5,32.5,4.1;3.2,2.9,0,1.5,3.3,35.7,1.2;3.9,2.1,1.5,0,3.6,34.5,2.6;...3.7,4.5,3.3,3.6,0,37,3.8;35,32.5,35.7,34.5,37,0,38.5;2,4.1,1.2,2.6,3.8,38.5,0];%距离矩阵t=[0,0.16,0.08,0.1,0.1,0.88,0.05;0.16,0,0.07,0.05,0.11,0.81,0.1;0.08,0.07,0,0.04,0.08,0.9,0.03;...0.1,0.05,0.04,0,0.09,0.86,0.07;0.1,0.11,0.08,0.09,0,0.92,0.1;0.88,0.81,0.9,0.86,0.92,0,0.96;...0.05,0.1,0.03,0.07,0.1,0.96,0;];%行驶时间矩阵w=[0.15,0.2,0.18,0.25,0.2,0.22];%服务时间矩阵%%时间窗向量early=[0.15,0.3,0.7,0.4,0.7,0.6];late=[2.5,3.4,3.3,2.7,2.5,4.5];xx=x;%取出染色体j=1;%分工点初始化%%取距离向量d1,d2d1=zeros(1,7);d1(1)=d(1,xx(1)+1);for i=1:5d1(i+1)=d(xx(i)+1,xx(i+1)+1);endd1(7)=d(xx(6)+1,1);%%时间窗计算T=t(1,xx(1)+1);pun1=0;if T<early(xx(1))pun1=early(xx(1))-T;T=early(xx(1));endT=T+w(xx(1));for i=2:6T=T+t(xx(i-1)+1,xx(i)+1);if T<early(xx(i))pun1=pun1+early(xx(i))-T;T=early(xx(i));endT=T+w(xx(6));endF=sum(10.*d1) +20*pun1;Fitness=1/F;计算时间文件：function[T]=TOTALT2(Xp1)Xp=Xp1;t=[0,0.16,0.08,0.1,0.1,0.88,0.05;0.16,0,0.07,0.05,0.11,0.81,0.1;0.08,0.07,0,0.04,0.08,0.9,0.03;...0.1,0.05,0.04,0,0.09,0.86,0.07;0.1,0.11,0.08,0.09,0,0.92,0.1;0.88,0.81,0.9,0.86,0.92,0,0.96;... 0.05,0.1,0.03,0.07,0.1,0.96,0;];%行驶时间矩阵w=[0.15,0.2,0.18,0.25,0.2,0.22];%服务时间矩阵%%时间窗向量early=[0.15,0.3,0.7,0.4,0.7,0.6];T=t(1,Xp(1)+1);if T<early(Xp(1))T=early(Xp(1));endT=T+w(Xp(1));for i=2:6T=T+t(Xp(i-1)+1,Xp(i)+1);if T<early(Xp(i))T=early(Xp(1));endT=T+w(Xp(i));endT=T+t(1,Xp(6)+1)。

MATLAB实现算法代码：GA（遗传算法）——整数编码function [BestGene,aa] = GA(MaxGeneration,GeneSize,GeneNum,pcross,pmute,minGene,maxGene)Parent = Init(GeneSize,GeneNum,minGene,maxGene);[BestGene,Parent] = KeepBest(Parent);aa = [];for i = 1:MaxGeneration[i 1/value(BestGene)]Child = chose(Parent);Child = cross(Child,pcross);Child = mute(Child,pmute,maxGene);[BestGene,Parent] = KeepBest(Child);aa = [aa;value(BestGene)];endfunction GeneInit = Init(GeneSize,GeneNum,minGene,maxGene)GeneInit = [];for i = 1:GeneSizex = []; x = ceil(rand(1,GeneNum).*(maxGene-minGene)) + minGene;GeneInit = [GeneInit;x];endGeneInit = [GeneInit;x];function Child = chose(Parent)GeneSize = size(Parent,1);for i = 1:GeneSizex = Parent(i,:);val(i) = value(x);endValSum = sum(val);val = val / ValSum;for i = 2:GeneSizeval(i) = val(i) + val(i-1);endfor i = 1:GeneSizerandval = rand;if randval <= val(1)Child(i,:) = Parent(1,:);endfor j = 2:GeneSizeif randval > val(j-1) && randval <= val(j)Child(i,:) = Parent(j,:);break;endendendChild(end,:) = Parent(end,:);function Child = cross(Parent,pcross)[GeneSize,GeneNum] = size(Parent);GeneSize = GeneSize - 1;Child = Parent;for i = 1:GeneSize/2if rand < pcrossflag = 0;while( flag==0 )randval1 = floor((GeneNum-1)*rand) + 1;randval2 = floor((GeneNum-1)*rand) + 1;if randval1 ~= randval2flag = 1;endendtemp = Child(2*i-1,randval1:randval2);Child(2*i-1,randval1:randval2) = Child(2*i,randval1:randval2);Child(2*i,randval1:randval2) = temp;endendfunction Child = mute(Parent,pmute,maxGene)[GeneSize,GeneNum] = size(Parent);GeneSize = GeneSize - 1;Child = Parent;for i = 1:GeneSizeif rand < pmuterandval = ceil((GeneNum-1)*rand) + 1;Child(i,randval) = maxGene(randval) - Child(i,randval) + 1;endendfunction [BestGene,Parent] = KeepBest(Child)[GeneSize,GeneNum] = size(Child);for i = 1:GeneSizex = Child(i,:);val(i) = value(x);endBigVal = val(1);flag = 1;for i = 2:GeneSizeif BigVal < val(i)BigVal = val(i);flag = i;endendBestGene = Child(flag,:); Parent = Child;Parent(1,:) = BestGene; Parent(end,:) = BestGene;。

function youhuafunD=code;N=50; % Tunablemaxgen=50; % Tunablecrossrate=0.5; %Tunablemuterate=0.08; %Tunablegeneration=1;num = length(D);fatherrand=randint(num,N,3);score = zeros(maxgen,N);while generation<=maxgenind=randperm(N-2)+2; % 随机配对交叉A=fatherrand(:,ind(1:(N-2)/2));B=fatherrand(:,ind((N-2)/2+1:end));% 多点交叉rnd=rand(num,(N-2)/2);ind=rnd tmp=A(ind);A(ind)=B(ind);B(ind)=tmp;% % 两点交叉% for kk=1:(N-2)/2% rndtmp=randint(1,1,num)+1;% tmp=A(1:rndtmp,kk);% A(1:rndtmp,kk)=B(1:rndtmp,kk);% B(1:rndtmp,kk)=tmp;% endfatherrand=[fatherrand(:,1:2),A,B];% 变异rnd=rand(num,N);ind=rnd [m,n]=size(ind);tmp=randint(m,n,2)+1;tmp(:,1:2)=0;fatherrand=tmp+fatherrand;fatherrand=mod(fatherrand,3);% fatherrand(ind)=tmp;%评价、选择scoreN=scorefun(fatherrand,D);% 求得N个个体的评价函数score(generation,:)=scoreN;[scoreSort,scoreind]=sort(scoreN);sumscore=cumsum(scoreSort);sumscore=sumscore./sumscore(end);childind(1:2)=scoreind(end-1:end);for k=3:Ntmprnd=rand;tmpind=tmprnd difind=[0,diff(tmpind)];if ~any(difind)difind(1)=1;endchildind(k)=scoreind(logical(difind));endfatherrand=fatherrand(:,childind);generation=generation+1;end% scoremaxV=max(score,[],2);minV=11*300-maxV;plot(minV,'*');title('各代的目标函数值');F4=D(:,4);FF4=F4-fatherrand(:,1);FF4=max(FF4,1);D(:,5)=FF4;save DData Dfunction D=codeload youhua.mat% properties F2 and F3F1=A(:,1);F2=A(:,2);F3=A(:,3);if (max(F2)>1450)||(min(F2)<=900)error('DATA property F2 exceed it''s range (900,1450]') end% get group property F1 of data, according to F2 value F4=zeros(size(F1));for ite=11:-1:1index=find(F2<=900+ite*50);F4(index)=ite;endD=[F1,F2,F3,F4];function ScoreN=scorefun(fatherrand,D)F3=D(:,3);F4=D(:,4);N=size(fatherrand,2);FF4=F4*ones(1,N);FF4rnd=FF4-fatherrand;FF4rnd=max(FF4rnd,1);ScoreN=ones(1,N)*300*11;% 这里有待优化for k=1:NFF4k=FF4rnd(:,k);for ite=1:11F0index=find(FF4k==ite);if ~isempty(F0index)tmpMat=F3(F0index);tmpSco=sum(tmpMat);ScoreBin(ite)=mod(tmpSco,300);endendScorek(k)=sum(ScoreBin);endScoreN=ScoreN-Scorek;遗传算法实例:% 下面举例说明遗传算法%% 求下列函数的最大值%% f(x)=10*sin(5x)+7*cos(4x) x∈[0,10] %% 将x 的值用一个10位的二值形式表示为二值问题，一个10位的二值数提供的分辨率是每为(10-0)/(2^10-1)≈0.01 。

遗传算法matlab代码以下是一个简单的遗传算法的MATLAB 代码示例：matlab复制代码% 遗传算法参数设置pop_size = 50; % 种群大小num_vars = 10; % 变量数目num_generations = 100; % 进化的代数mutation_rate = 0.01; % 变异率crossover_rate = 0.8; % 交叉率% 初始化种群population = rand(pop_size, num_vars);% 开始进化for i = 1:num_generations% 计算适应度fitness = evaluate_fitness(population);% 选择操作selected_population = selection(population, fitness);% 交叉操作offspring_population = crossover(selected_population,crossover_rate);% 变异操作mutated_population = mutation(offspring_population,mutation_rate);% 生成新种群population = [selected_population; mutated_population];end% 选择最优解best_solution = population(find(fitness == max(fitness)), :);% 适应度函数function f = evaluate_fitness(population)f = zeros(size(population));for i = 1:size(population, 1)f(i) = sum(population(i, :));endend% 选择函数function selected_population = selection(population, fitness)% 轮盘赌选择total_fitness = sum(fitness);probabilities = fitness / total_fitness;selected_indices = zeros(pop_size, 1);for i = 1:pop_sizer = rand();cumulative_probabilities = cumsum(probabilities);for j = 1:pop_sizeif r <= cumulative_probabilities(j)selected_indices(i) = j;break;endendendselected_population = population(selected_indices, :);end% 交叉函数function offspring_population = crossover(parental_population, crossover_rate)offspring_population = zeros(size(parental_population));num_crossovers = ceil(size(parental_population, 1) *crossover_rate);crossover_indices = randperm(size(parental_population, 1),num_crossovers);以下是另一个一个简单的遗传算法的MATLAB 代码示例：matlab复制代码% 初始化种群population = rand(nPopulation, nGenes);% 进化迭代for iGeneration = 1:nGeneration% 计算适应度fitness = evaluateFitness(population);% 选择父代parentIdx = selection(fitness);parent = population(parentIdx, :);% 交叉产生子代child = crossover(parent);% 变异子代child = mutation(child);% 更新种群population = [parent; child];end% 评估最优解bestFitness = -Inf;for i = 1:nPopulationf = evaluateFitness(population(i, :));if f > bestFitnessbestFitness = f;bestIndividual = population(i, :);endend% 可视化结果plotFitness(fitness);其中，nPopulation和nGenes分别是种群大小和基因数；nGeneration是迭代次数；evaluateFitness函数用于计算个体的适应度；selection函数用于选择父代；crossover函数用于交叉产生子代；mutation函数用于变异子代。

附页：一．遗传算法源程序：clc; clear;population;％评价目标函数值for uim=1：popsizevector=population(uim,:）；obj(uim）=hanshu(hromlength，vector，phen);end％obj%min（obj）clear uim;objmin=min（obj）;for sequ=1：popsizeif obj(sequ）==objminopti=population（sequ,：)；endendclear sequ；fmax=22000；%==for gen=1：maxgen%选择操作%将求最小值的函数转化为适应度函数for indivi=1：popsizeobj1（indivi)=1/obj（indivi)；endclear indivi；％适应度函数累加总合total=0;for indivi=1：popsizetotal=total+obj1(indivi)；endclear indivi;%每条染色体被选中的几率for indivi=1:popsizefitness1（indivi）=obj1(indivi）/total；endclear indivi;%各条染色体被选中的范围for indivi=1：popsizefitness(indivi)=0；for j=1:indivifitness(indivi）=fitness(indivi)+fitness1(j）；endendclear j；fitness;%选择适应度高的个体for ranseti=1：popsizeran=rand；while (ran>1|｜ran<0)ran=rand;endran;if ran〈=fitness(1）newpopulation(ranseti,：)=population(1，:）;elsefor fet=2:popsizeif (ran〉fitness(fet—1))&&(ran<=fitness(fet)）newpopulation(ranseti，：)=population(fet，：)；endendendendclear ran;newpopulation；％交叉for int=1:2：popsize-1popmoth=newpopulation（int,：）；popfath=newpopulation（int+1，:）；popcross(int，:）=popmoth；popcross（int+1，：)=popfath；randnum=rand；if(randnum〈 P>cpoint1=round（rand＊hromlength）；cpoint2=round(rand*hromlength)；while （cpoint2==cpoint1）cpoint2=round（rand*hromlength);endif cpoint1>cpoint2tem=cpoint1；cpoint1=cpoint2；cpoint2=tem;endcpoint1；cpoint2;for term=cpoint1+1:cpoint2for ss=1:hromlengthif popcross(int，ss)==popfath（term）tem1=popcross（int,ss)；popcross（int，ss）=popcross(int,term);popcross(int,term）=tem1；endendclear tem1；endfor term=cpoint1+1：cpoint2for ss=1：hromlengthif popcross(int+1，ss)==popmoth（term)tem1=popcross（int+1,ss）;popcross（int+1,ss)=popcross（int+1,term）;popcross（int+1,term）=tem1;endendclear tem1；endendclear term;endclear randnum;popcross;%变异操作newpop=popcross；for int=1:popsizerandnum=rand;if randnumcpoint12=round（rand*hromlength)；cpoint22=round(rand*hromlength)；if （cpoint12==0）cpoint12=1；endif (cpoint22==0）cpoint22=1;endwhile (cpoint22==cpoint12)cpoint22=round（rand＊hromlength)；if cpoint22==0;cpoint22=1;endendtemp=newpop(int，cpoint12）；newpop(int,cpoint12)=newpop(int,cpoint22)；newpop(int,cpoint22)=temp;。

遗传算法Matlab源代码完整可以运行的数值优化遗传算法源代码function[X,MaxFval,BestPop,Trace]=fga(FUN,bounds,MaxEranum,PopSiz e,options,pCross,pMutation,pInversion)%[X,MaxFval,BestPop,Trace]=fga(FUN,bounds,MaxEranum,PopSiz e,options,pCross,pMutation,pInversion)% Finds a maximum of a function of several variables.% fga solves problems of the form:% max F(X) subject to: LB = X = UB (LB=bounds(:,1),UB=bounds(:,2))% X - 最优个体对应自变量值% MaxFval - 最优个体对应函数值% BestPop - 最优的群体即为最优的染色体群% Trace - 每代最佳个体所对应的目标函数值% FUN - 目标函数% bounds - 自变量范围% MaxEranum - 种群的代数,取50--500(默认200)% PopSize - 每一代种群的规模；此可取50--200(默认100)% pCross - 交叉概率,一般取0.5--0.85之间较好(默认0.8)% pMutation - 初始变异概率,一般取0.05-0.2之间较好(默认0.1)% pInversion - 倒位概率,一般取0.05－0.3之间较好(默认0.2) % options - 1*2矩阵,options(1)=0二进制编码(默认0),option(1)~=0十进制编码,option(2)设定求解精度(默认1e-4)T1=clock;%检验初始参数if nargin2, error('FMAXGA requires at least three input arguments'); endif nargin==2, MaxEranum=150;PopSize=100;options=[1 1e-4];pCross=0.85;pMutation=0.1;pInversion=0.25;endif nargin==3, PopSize=100;options=[1 1e-4];pCross=0.85;pMutation=0.1;pInversion=0.25;endif nargin==4, options=[1 1e-4];pCross=0.85;pMutation=0.1;pInversion=0.25;endif nargin==5, pCross=0.85;pMutation=0.1;pInversion=0.25;endif nargin==6, pMutation=0.1;pInversion=0.25;endif nargin==7, pInversion=0.25;endif (options(1)==0|options(1)==1)find((bounds(:,1)-bounds(:,2))0)error('数据输入错误,请重新输入:');end% 定义全局变量global m n NewPop children1 children2 VarNum% 初始化种群和变量precision = options(2);bits = ceil(log2((bounds(:,2)-bounds(:,1))' ./ precision));%由设定精度划分区间VarNum = size(bounds,1);[Pop] = InitPop(PopSize,bounds,bits,options);%初始化种群[m,n] = size(Pop);fit = zeros(1,m);NewPop = zeros(m,n);children1 = zeros(1,n);children2 = zeros(1,n);pm0 = pMutation;BestPop = zeros(MaxEranum,n);%分配初始解空间BestPop,TraceTrace = zeros(1,MaxEranum);完整可以运行的数值优化遗传算法源代码Lb = ones(PopSize,1)*bounds(:,1)';Ub = ones(PopSize,1)*bounds(:,2)';%二进制编码采用多点交叉和均匀交叉，并逐步增大均匀交叉概率%浮点编码采用离散交叉（前期）、算术交叉（中期）、AEA重组（后期）OptsCrossOver = [ones(1,MaxEranum)*options(1);...round(unidrnd(2*(MaxEranum-[1:MaxEranum]))/MaxEranum)]';%浮点编码时采用两种自适应变异和一种随机变异（自适应变异发生概率为随机变异发生的2倍）OptsMutation = [ones(1,MaxEranum)*options(1);unidrnd(5,1,MaxEranum)]';if options(1)==3D=zeros(n);CityPosition=bounds;D = sqrt((CityPosition(:, ones(1,n)) - CityPosition(:, ones(1,n))').^2 +...(CityPosition(:,2*ones(1,n)) - CityPosition(:,2*ones(1,n))').^2 );end%========================================================================== % 进化主程序%%===================================== ===================================== eranum = 1;H=waitbar(0,'Please wait...');while(eranum=MaxEranum)for j=1:mif options(1)==1%eval(['[fit(j)]=' FUN '(Pop(j,:));']);%但执行字符串速度比直接计算函数值慢fit(j)=feval(FUN,Pop(j,:));%计算适应度elseif options(1)==0%eval(['[fit(j)]=' FUN '(b2f(Pop(j,:),bounds,bits));']);fit(j)=feval(FUN,(b2f(Pop(j,:),bounds,bits)));elsefit(j)=-feval(FUN,Pop(j,:),D);endend[Maxfit,fitIn]=max(fit);%得到每一代最大适应值Meanfit(eranum)=mean(fit);BestPop(eranum,:)=Pop(fitIn,:);Trace(eranum)=Maxfit;if options(1)==1Pop=(Pop-Lb)./(Ub-Lb);%将定义域映射到[0,1]：[Lb,Ub]--[0,1] ,Pop--(Pop-Lb)./(Ub-Lb)endswitch round(unifrnd(0,eranum/MaxEranum))%进化前期尽量使用实行锦标赛选择，后期逐步增大非线性排名选择case {0} [selectpop]=TournamentSelect(Pop,fit,bits);%锦标赛选择case {1}[selectpop]=NonlinearRankSelect(Pop,fit,bits);%非线性排名选择end完整可以运行的数值优化遗传算法源代码[CrossOverPop]=CrossOver(selectpop,pCross,OptsCrossOver(er anum,:));%交叉[MutationPop]=Mutation(CrossOverPop,fit,pMutation,VarNum,O ptsMutation(eranum,:)); %变异[InversionPop]=Inversion(MutationPop,pInversion);%倒位%更新种群if options(1)==1Pop=Lb+InversionPop.*(Ub-Lb);%还原PopelsePop=InversionPop;endpMutation=pm0+(eranum^3)*(pCross/2-pm0)/(eranum^4); %逐步增大变异率至1/2交叉率percent=num2str(round(100*eranum/MaxEranum));waitbar(eranum/MaxEranum,H,['Evolution complete ',percent,'%']);eranum=eranum+1;endclose(H);% 格式化输出进化结果和解的变化情况t=1:MaxEranum;plot(t,Trace,t,Meanfit);legend('解的变化','种群的变化');title('函数优化的遗传算法');xlabel('进化世代数');ylabel('每一代最优适应度');[MaxFval,MaxFvalIn]=max(Trace);if options(1)==1|options(1)==3X=BestPop(MaxFvalIn,:);elseif options(1)==0X=b2f(BestPop(MaxFvalIn,:),bounds,bits);endhold on;plot(MaxFvalIn,MaxFval,'*');text(MaxFvalIn+5,MaxFval,['FMAX=' num2str(MaxFval)]);str1=sprintf(' Best generation:\n %d\n\n Best X:\n %s\n\n MaxFval\n %f\n',...MaxFvalIn,num2str(X),MaxFval);disp(str1);% -计时T2=clock;elapsed_time=T2-T1;if elapsed_time(6)0elapsed_time(6)=elapsed_time(6)+60;elapsed_time(5)=elapsed_time(5)-1;endif elapsed_time(5)0elapsed_time(5)=elapsed_time(5)+60;elapsed_time(4)=elapsed_t ime(4)-1;end完整可以运行的数值优化遗传算法源代码str2=sprintf('elapsed_time\n %d (h) %d (m) %.4f (s)',elapsed_time(4),elapsed_time(5),elapsed_time(6));disp(str2);%===================================== ===================================== % 遗传操作子程序%%===================================== ===================================== % -- 初始化种群--% 采用浮点编码和二进制Gray编码(为了克服二进制编码的Hamming悬崖缺点)function [initpop]=InitPop(popsize,bounds,bits,options)numVars=size(bounds,1);%变量数目rang=(bounds(:,2)-bounds(:,1))';%变量范围if options(1)==1initpop=zeros(popsize,numVars);initpop=(ones(popsize,1)*rang).*(rand(popsize,numVars))+(ones (popsize,1)*bounds(:,1)');elseif options(1)==0precision=options(2);%由求解精度确定二进制编码长度len=sum(bits);initpop=zeros(popsize,len);%The whole zero encoding individualfor i=2:popsize-1pop=round(rand(1,len));pop=mod(([0 pop]+[pop 0]),2);%i=1时,b(1)=a(1);i1时,b(i)=mod(a(i-1)+a(i),2)%其中原二进制串:a(1)a(2)...a(n),Gray串:b(1)b(2)...b(n)initpop(i,:)=pop(1:end-1);endinitpop(popsize,:)=ones(1,len);%The whole one encoding individualelsefor i=1:popsizeinitpop(i,:)=randperm(numVars);%为Tsp问题初始化种群endend% -- 二进制串解码--function [fval] = b2f(bval,bounds,bits)% fval - 表征各变量的十进制数% bval - 表征各变量的二进制编码串% bounds - 各变量的取值范围% bits - 各变量的二进制编码长度scale=(bounds(:,2)-bounds(:,1))'./(2.^bits-1); %The range of the variablesnumV=size(bounds,1);cs=[0 cumsum(bits)];for i=1:numVa=bval((cs(i)+1):cs(i+1));fval(i)=sum(2.^(size(a,2)-1:-1:0).*a)*scale(i)+bounds(i,1);end% -- 选择操作--完整可以运行的数值优化遗传算法源代码% 采用基于轮盘赌法的非线性排名选择% 各个体成员按适应值从大到小分配选择概率：% P(i)=(q/1-(1-q)^n)*(1-q)^i, 其中P(0)P(1)...P(n), sum(P(i))=1function [NewPop]=NonlinearRankSelect(OldPop,fit,bits) global m n NewPopfit=fit';selectprob=fit/sum(fit);%计算各个体相对适应度(0,1)q=max(selectprob);%选择最优的概率x=zeros(m,2);x(:,1)=[m:-1:1]';[y x(:,2)]=sort(selectprob);r=q/(1-(1-q)^m);%标准分布基值newfit(x(:,2))=r*(1-q).^(x(:,1)-1);%生成选择概率newfit=[0 cumsum(newfit)];%计算各选择概率之和rNums=rand(m,1);newIn=1;while(newIn=m)NewPop(newIn,:)=OldPop(length(find(rNums(newIn)newfit)),:);newIn=newIn+1;end% -- 锦标赛选择(含精英选择) --function [NewPop]=TournamentSelect(OldPop,fit,bits)global m n NewPopnum=floor(m./2.^(1:10));num(find(num==0))=[];L=length(num);a=sum(num);b=m-a;PopIn=1;while(PopIn=L)r=unidrnd(m,num(PopIn),2^PopIn);[LocalMaxfit,In]=max(fit(r),[],2);SelectIn=r((In-1)*num(PopIn)+[1:num(PopIn)]');NewPop(sum(num(1:PopIn))-num(PopIn)+1:sum(num(1:PopIn)),:)=OldPop(SelectIn,:);PopIn=PopIn+1;r=[];In=[];LocalMaxfit=[];endif b1NewPop((sum(num)+1):(sum(num)+b-1),:)=OldPop(unidrnd(m,1,b-1),:);end[GlobalMaxfit,I]=max(fit);%保留每一代中最佳个体NewPop(end,:)=OldPop(I,:);% -- 交叉操作--function [NewPop]=CrossOver(OldPop,pCross,opts)global m n NewPopr=rand(1,m);完整可以运行的数值优化遗传算法源代码y1=find(rpCross);y2=find(r=pCross);len=length(y1);if len==1|(len2mod(len,2)==1)%如果用来进行交叉的染色体的条数为奇数，将其调整为偶数y2(length(y2)+1)=y1(len);y1(len)=[];endi=0;if length(y1)=2if opts(1)==1%浮点编码交叉while(i=length(y1)-2)NewPop(y1(i+1),:)=OldPop(y1(i+1),:);NewPop(y1(i+2),:)=OldPop(y1(i+2),:);if opts(2)==0n1%discret crossoverPoints=sort(unidrnd(n,1,2));NewPop(y1(i+1),Points(1):Points(2))=OldPop(y1(i+2),Points(1):Po ints(2));NewPop(y1(i+2),Points(1):Points(2))=OldPop(y1(i+1),Points(1):Po ints(2));elseif opts(2)==1%arithmetical crossoverPoints=round(unifrnd(0,pCross,1,n));CrossPoints=find(Points==1);r=rand(1,length(CrossPoints));NewPop(y1(i+1),CrossPoints)=r.*OldPop(y1(i+1),CrossPoints)+(1 -r).*OldPop(y1(i+2),CrossPoints);NewPop(y1(i+2),CrossPoints)=r.*OldPop(y1(i+2),CrossPoints)+(1 -r).*OldPop(y1(i+1),CrossPoints); else %AEA recombination Points=round(unifrnd(0,pCross,1,n));CrossPoints=find(Points==1);v=unidrnd(4,1,2);NewPop(y1(i+1),CrossPoints)=(floor(10^v(1)*OldPop(y1(i+1),Cro ssPoints))+...10^v(1)*OldPop(y1(i+2),CrossPoints)-floor(10^v(1)*OldPop(y1(i+2),CrossPoints)))/10^v(1);NewPop(y1(i+2),CrossPoints)=(floor(10^v(2)*OldPop(y1(i+2),Cro ssPoints))+...10^v(2)*OldPop(y1(i+1),CrossPoints)-floor(10^v(2)*OldPop(y1(i+1),CrossPoints)))/10^v(2);endi=i+2;endelseif opts(1)==0%二进制编码交叉while(i=length(y1)-2)if opts(2)==0[NewPop(y1(i+1),:),NewPop(y1(i+2),:)]=EqualCrossOver(OldPop( y1(i+1),:),OldPop(y1(i+2),:)); else[NewPop(y1(i+1),:),NewPop(y1(i+2),:)]=MultiPointCross(OldPop( y1(i+1),:),OldPop(y1(i+2),:)); endi=i+2;endelse %Tsp问题次序杂交for i=0:2:length(y1)-2xPoints=sort(unidrnd(n,1,2));NewPop([y1(i+1)y1(i+2)],xPoints(1):xPoints(2))=OldPop([y1(i+2)y1(i+1)],xPoints(1):xPoints(2));完整可以运行的数值优化遗传算法源代码%NewPop(y1(i+2),xPoints(1):xPoints(2))=OldPop(y1(i+1),xPo ints(1):xPoints(2));temp=[OldPop(y1(i+1),xPoints(2)+1:n)OldPop(y1(i+1),1:xPoints(2))];for del1i=xPoints(1):xPoints(2)temp(find(temp==OldPop(y1(i+2),del1i)))=[];endNewPop(y1(i+1),(xPoints(2)+1):n)=temp(1:(n-xPoints(2)));NewPop(y1(i+1),1:(xPoints(1)-1))=temp((n-xPoints(2)+1):end);temp=[OldPop(y1(i+2),xPoints(2)+1:n)OldPop(y1(i+2),1:xPoints(2))];for del2i=xPoints(1):xPoints(2)temp(find(temp==OldPop(y1(i+1),del2i)))=[];endNewPop(y1(i+2),(xPoints(2)+1):n)=temp(1:(n-xPoints(2)));NewPop(y1(i+2),1:(xPoints(1)-1))=temp((n-xPoints(2)+1):end);endendendNewPop(y2,:)=OldPop(y2,:);% -二进制串均匀交叉算子function[children1,children2]=EqualCrossOver(parent1,parent2) global n children1 children2hidecode=round(rand(1,n));%随机生成掩码crossposition=find(hidecode==1);holdposition=find(hidecode==0);children1(crossposition)=parent1(crossposition);%掩码为1，父1为子1提供基因children1(holdposition)=parent2(holdposition);%掩码为0，父2为子1提供基因children2(crossposition)=parent2(crossposition);%掩码为1，父2为子2提供基因children2(holdposition)=parent1(holdposition);%掩码为0，父1为子2提供基因% -二进制串多点交叉算子function[Children1,Children2]=MultiPointCross(Parent1,Parent2)%交叉点数由变量数决定global n Children1 Children2 VarNumChildren1=Parent1;Children2=Parent2;Points=sort(unidrnd(n,1,2*VarNum));for i=1:VarNumChildren1(Points(2*i-1):Points(2*i))=Parent2(Points(2*i-1):Points(2*i));Children2(Points(2*i-1):Points(2*i))=Parent1(Points(2*i-1):Points(2*i));end% -- 变异操作--function[NewPop]=Mutation(OldPop,fit,pMutation,VarNum,opts) global m n NewPopNewPop=OldPop;r=rand(1,m);MutIn=find(r=pMutation);L=length(MutIn);完整可以运行的数值优化遗传算法源代码i=1;if opts(1)==1%浮点变异maxfit=max(fit);upfit=maxfit+0.05*abs(maxfit);if opts(2)==1|opts(2)==3while(i=L)%自适应变异（自增或自减）Point=unidrnd(n);T=(1-fit(MutIn(i))/upfit)^2;q=abs(1-rand^T);%if q1%按严格数学推理来说，这段程序是不能缺少的% q=1%endp=OldPop(MutIn(i),Point)*(1-q);if unidrnd(2)==1NewPop(MutIn(i),Point)=p+q;elseNewPop(MutIn(i),Point)=p;endi=i+1;endelseif opts(2)==2|opts(2)==4%AEA变异（任意变量的某一位变异）while(i=L)Point=unidrnd(n);T=(1-abs(upfit-fit(MutIn(i)))/upfit)^2;v=1+unidrnd(1+ceil(10*T));%v=1+unidrnd(5+ceil(10*eranum/MaxEranum));q=mod(floor(OldPop(MutIn(i),Point)*10^v),10);NewPop(MutIn(i),Point)=OldPop(MutIn(i),Point)-(q-unidrnd(9))/10^v;i=i+1;endelsewhile(i=L)Point=unidrnd(n);if round(rand)NewPop(MutIn(i),Point)=OldPop(MutIn(i),Point)*(1-rand);elseNewPop(MutIn(i),Point)=OldPop(MutIn(i),Point)+(1-OldPop(MutIn(i),Point))*rand; endi=i+1;endendelseif opts(1)==0%二进制串变异if L=1while i=Lk=unidrnd(n,1,VarNum); %设置变异点数(=变量数)for j=1:length(k)if NewPop(MutIn(i),k(j))==1NewPop(MutIn(i),k(j))=0;else完整可以运行的数值优化遗传算法源代码NewPop(MutIn(i),k(j))=1;endendi=i+1;endendelse%Tsp变异if opts(2)==1|opts(2)==2|opts(2)==3|opts(2)==4numMut=ceil(pMutation*m);r=unidrnd(m,numMut,2);[LocalMinfit,In]=min(fit(r),[],2);SelectIn=r((In-1)*numMut+[1:numMut]');while(i=numMut)mPoints=sort(unidrnd(n,1,2));if mPoints(1)~=mPoints(2)NewPop(SelectIn(i),1:mPoints(1)-1)=OldPop(SelectIn(i),1:mPoints(1)-1);NewPop(SelectIn(i),mPoints(1):mPoints(2)-1)=OldPop(SelectIn(i),mPoints(1)+1:mPoints(2));NewPop(SelectIn(i),mPoints(2))=OldPop(SelectIn(i),mPoints(1));NewPop(SelectIn(i),mPoints(2)+1:n)=OldPop(SelectIn(i),mPoints( 2)+1:n);elseNewPop(SelectIn(i),:)=OldPop(SelectIn(i),:);endi=i+1;endr=rand(1,m);MutIn=find(r=pMutation);L=length(MutIn);while i=LmPoints=sort(unidrnd(n,1,2));rIn=randperm(mPoints(2)-mPoints(1)+1);NewPop(MutIn(i),mPoints(1):mPoints(2))=OldPop(MutIn(i),mPoin ts(1)+rIn-1);i=i+1;endendend% -- 倒位操作--function [NewPop]=Inversion(OldPop,pInversion)global m n NewPopNewPop=OldPop;r=rand(1,m);PopIn=find(r=pInversion);len=length(PopIn);if len=1while(i=len)d=sort(unidrnd(n,1,2));完整可以运行的数值优化遗传算法源代码NewPop(PopIn(i),d(1):d(2))=OldPop(PopIn(i),d(2):-1:d(1)); i=i+1;。

顺序选择遗传算法MATLAB代码function [xv,fv] = SBOGA(fitness,a,b,NP,NG,q,Pc,Pm,eps)%顺序选择遗传算法L = ceil(log2((b-a)/eps+1)); %根据离散精度，确定二进制编码需要的码长x = zeros(NP,L);for i=1:NPx(i,:) = Initial(L); %种群初始化fx(i) = fitness(Dec(a,b,x(i,:),L)); %个体适应值endfor k=1:NG[sortf,sortx] = sort(fx); %适应值排序x = x(sortx,:);fx = fx(sortx);for i=1:NP %固定选择概率Px(i) = (1-q)^(NP-i)*q/(1-(1-q)^NP);endPPx = 0;PPx(1) = Px(1);for i=2:NP %用于轮盘赌策略的概率累加PPx(i) = PPx(i-1) + Px(i);endfor i=1:NPsita = rand();for n=1:NPif sita <= PPx(n)SelFather = n; %根据轮盘赌策略确定的父亲break;endendSelmother = floor(rand()*(NP-1))+1; %随机选择母亲posCut = floor(rand()*(L-2)) + 1; %随机确定交叉点r1 = rand();if r1<=Pc %交叉nx(i,1:posCut) = x(SelFather,1:posCut); nx(i,(posCut+1):L) = x(Selmother,(posCut+1):L);r2 = rand();if r2 <= Pm %变异posMut = round(rand()*(L-1) + 1);nx(i,posMut) = ~nx(i,posMut);endelsenx(i,:) = x(SelFather,:);endendx = nx;for i=1:NPfx(i) = fitness(Dec(a,b,x(i,:),L)); %子代适应值endendfv = -inf;for i=1:NPfitx = fitness(Dec(a,b,x(i,:),L));if fitx > fvfv = fitx; %取个体中的最好值作为最终结果xv = Dec(a,b,x(i,:),L);endendfunction result = Initial(length) %初始化函数for i=1:lengthr = rand();result(i) = round(r);endfunction y = Dec(a,b,x,L) %二进制编码转换为十进制编码base = 2.^((L-1):-1:0);y = dot(base,x);y = a + y*(b-a)/(2^L-1);。

遗传算法优缺点遗传算法的优点：1. 与问题领域无关切快速随机的搜索能力。

2. 搜索从群体出发，具有潜在的并行性，可以进行多个个体的同时比较，robust.3. 搜索使用评价函数启发，过程简单4. 使用概率机制进行迭代，具有随机性。

5. 具有可扩展性，容易与其他算法结合。

缺点是：1。

没有能够及时利用网络的反馈信息,故算法的搜索速度比较慢，要得要较精确的解需要较多的训练时间。

2。

算法对初始种群的选择有一定的依赖性，能够结合一些启发算法进行改进。

3。

算法的并行机制的潜在能力没有得到充分的利用，这也是当前遗传算法的一个研究热点方向。

核心函数：(1)function [pop]=initializega(num,bounds,eevalFN,eevalOps,options)--初始种群的生成函数【输出参数】pop--生成的初始种群【输入参数】num--种群中的个体数目bounds--代表变量的上下界的矩阵eevalFN--适应度函数eevalOps--传递给适应度函数的参数options--选择编码形式(浮点编码或是二进制编码)[precision F_or_B],如precision--变量进行二进制编码时指定的精度F_or_B--为1时选择浮点编码，否则为二进制编码,由precision指定精度)(2)function [x,endPop,bPop,traceInfo] = ga(bounds,evalFN,evalOps,startPop,opts,...termFN,termOps,selectFN,selectOps,xOverFNs,xOverOps,mutFNs,mutOps)--遗传算法函数【输出参数】x--求得的最优解endPop--最终得到的种群bPop--最优种群的一个搜索轨迹【输入参数】bounds--代表变量上下界的矩阵evalFN--适应度函数evalOps--传递给适应度函数的参数startPop-初始种群opts[epsilon prob_ops display]--opts(1:2)等同于initializega的options参数，第三个参数控制是否输出，一般为0。

matlab智能算法代码MATLAB是一种功能强大的数值计算和科学编程软件，它提供了许多智能算法的实现。

下面是一些常见的智能算法及其在MATLAB中的代码示例：1. 遗传算法（Genetic Algorithm）：MATLAB中有一个专门的工具箱，称为Global Optimization Toolbox，其中包含了遗传算法的实现。

以下是一个简单的遗传算法示例代码：matlab.% 定义目标函数。

fitness = @(x) x^2;% 设置遗传算法参数。

options = gaoptimset('Display', 'iter','PopulationSize', 50);% 运行遗传算法。

[x, fval] = ga(fitness, 1, options);2. 粒子群优化算法（Particle Swarm Optimization）：MATLAB中也有一个工具箱，称为Global Optimization Toolbox，其中包含了粒子群优化算法的实现。

以下是一个简单的粒子群优化算法示例代码：matlab.% 定义目标函数。

fitness = @(x) x^2;% 设置粒子群优化算法参数。

options = optimoptions('particleswarm', 'Display','iter', 'SwarmSize', 50);% 运行粒子群优化算法。

[x, fval] = particleswarm(fitness, 1, [], [], options);3. 支持向量机（Support Vector Machine）：MATLAB中有一个机器学习工具箱，称为Statistics and Machine Learning Toolbox，其中包含了支持向量机的实现。

function [R,Rlength]= GA_TSP(xyCity,dCity,Population,nPopulation,pCrossover,percent,pMutation,generation,nR,rr,rang eCity,rR,moffspring,record,pi,Shock,maxShock)clear allA=load('d.txt');AxyCity=[A(1,:);A(2,:)]; %x,y为各地点坐标xyCityfigure(1)grid onhold onscatter(xyCity(1,:),xyCity(2,:),'b+')grid onnCity=50;nCityfor i=1:nCity %计算城市间距离for j=1:nCitydCity(i,j)=abs(xyCity(1,i)-xyCity(1,j))+abs(xyCity(2,i)-xyCity(2,j));endend %计算城市间距离xyCity; %显示城市坐标dCity %显示城市距离矩阵%初始种群k=input('取点操作结束'); %取点时对操作保护disp('-------------------')nPopulation=input('种群个体数量：'); %输入种群个体数量if size(nPopulation,1)==0nPopulation=50; %默认值endfor i=1:nPopulationPopulation(i,:)=randperm(nCity-1); %产生随机个体endPopulation %显示初始种群pCrossover=input('交叉概率：'); %输入交叉概率percent=input('交叉部分占整体的百分比：'); %输入交叉比率pMutation=input('突变概率：'); %输入突变概率nRemain=input('最优个体保留最大数量：');pi(1)=input('选择操作最优个体被保护概率：');%输入最优个体被保护概率pi(2)=input('交叉操作最优个体被保护概率：');pi(3)=input('突变操作最优个体被保护概率：');maxShock=input('最大突变概率：');if size(pCrossover,1)==0pCrossover=0.85;endif size(percent,1)==0percent=0.5;endif size(pMutation,1)==0pMutation=0.05;endShock=0;rr=0;Rlength=0;counter1=0;counter2=0;R=zeros(1,nCity-1);[newPopulation,R,Rlength,counter2,rr]=select(Population,nPopulation,nCity,dCity,Rlength,R,coun ter2,pi,nRemain);R0=R;record(1,:)=R;rR(1)=Rlength;Rlength0=Rlength;generation=input('算法终止条件A.最多迭代次数：');%输入算法终止条件if size(generation,1)==0generation=200;endnR=input('算法终止条件B.最短路径连续保持不变代数：');if size(nR,1)==0nR=10;endwhile counter1<generation&counter2<nRif counter2<nR*1/5Shock=0;elseif counter2<nR*2/5Shock=maxShock*1/4-pMutation;elseif counter2<nR*3/5Shock=maxShock*2/4-pMutation;elseif counter2<nR*4/5Shock=maxShock*3/4-pMutation;elseShock=maxShock-pMutation;endcounter1newPopulationoffspring=crossover(newPopulation,nCity,pCrossover,percent,nPopulation,rr,pi,nRemain);offspringmoffspring=Mutation(offspring,nCity,pMutation,nPopulation,rr,pi,nRemain,Shock);[newPopulation,R,Rlength,counter2,rr]=select(moffspring,nPopulation,nCity,dCity,Rlength,R,coun ter2,pi,nRemain);counter1=counter1+1;rR(counter1+1)=Rlength;record(counter1+1,:)=R;endR0;Rlength0;R;Rlength;minR=min(rR);disp('最短路经出现代数:')rr=find(rR==minR)disp('最短路经:')record(rr,:);mR=record(rr(1,1),:)disp('终止条件一:')counter1disp('终止条件二:')counter2disp('最短路经长度:')minRdisp('最初路经长度:')rR(1)figure(2)plotaiwa(xyCity,mR,nCity)figure(3)i=1:counter1+1;plot(i,rR(i))grid onfunction[newPopulation,R,Rlength,counter2,rr]=select(Population,nPopulation,nCity,dCity,Rlength,R,coun ter2,pi,nRemain)Distance=zeros(nPopulation,1); %零化路径长度Fitness=zeros(nPopulation,1); %零化适应概率Sum=0; %路径长度for i=1:nPopulation %计算个体路径长度for j=1:nCity-2Distance(i)=Distance(i)+dCity(Population(i,j),Population(i,j+1));end %对路径长度调整，增加起始点到路径首尾点的距离Distance(i)=Distance(i)+dCity(Population(i,1),nCity)+dCity(Population(i,nCity-1),nCity);Sum=Sum+Distance(i); %累计总路径长度end %计算个体路径长度if Rlength==min(Distance)counter2=counter2+1;elsecounter2=0;endRlength=min(Distance); %更新最短路径长度Rlength;rr=find(Distance==Rlength);R=Population(rr(1,1),:); %更新最短路径for i=1:nPopulationFitness(i)=(max(Distance)-Distance(i)+0.001)/(nPopulation*(max(Distance)+0.001)-Sum); %适应概率=个体/总和。