GA搜索（浮点编码形式的 遗传算法 ）

一、原理：

原理可转至：遗传算法原理

二、待解决问题：

搜索下列函数的最小值：
$$\begin{gathered} f(x_1,x_2)=2\cos (x_1{x}_2)+4x_1+x_2^2 \\ {x}_1,x_2\in [-5,5] \end{gathered}$$

三、代码实现

clc
clear
close all

%% 采用浮点数编码形式搜索

N = 100;                                                        % 群体数量
ChromNum = 2;                                                   % 染色体数量
IterMax = 20;                                                   % 迭代次数
Pc = 0.7;                                                       % 交叉概率
Pm = 0.1;                                                       % 变异概率
UpLimit = [5,5];                                                % 染色体取值上限
DownLimit = [-5,-5];                                            % 染色体取值下限
WaitbarInter = IterMax / 100;                                   % 一个和进度条有关的参数
FitnessBestProgress = zeros(1,IterMax);                         % 初始化群体最优适应度变化情况存储容器
AveFitnessProgress = zeros(1,IterMax);                          % 初始化群体平均适应度变化情况存储容器
Individual = InitIndividual(N,ChromNum,DownLimit,UpLimit);      % 初始化群体染色体
fitness = CalcFitness(Individual);                              % 初始化适应度
[IndividualBest,FitnessBest] = FindBest(Individual,fitness);    % 初始化最优个体
FitnessBestProgress(1,1) = FitnessBest;                         % 存储当前群体最优适应度
AveFitnessProgress(1,1) = CalcAveFitness(fitness);              % 存储当前群体平均适应度

tic
h = waitbar(0, ['已完成:0%   运算中...用时:', num2str(toc)]);    % 初始化进度条 

for iter = 2:IterMax
    Individual = MutIndividual(Individual,iter,IterMax,Pm,DownLimit,UpLimit);   % 变异
    Individual = CroIndividual(Individual,Pc);                                  % 交叉
    fitness = CalcFitness(Individual);                                          % 计算当前群体适应度
    [IndividualBesttemp,FitnessBesttemp] = FindBest(Individual,fitness);        % 寻找当前群体最优个体
    if FitnessBesttemp < FitnessBest                                            % 最有个体存储迭代
        FitnessBest = FitnessBesttemp;
        IndividualBest = IndividualBesttemp;
    end
    [Individual,fitness] = NaturalSlection(Individual,fitness,IndividualBest,FitnessBest);  % 自然选择
    FitnessBestProgress(1,iter) = FitnessBest;                                  % 存储当前群体最优适应度
    AveFitnessProgress(1,iter) = CalcAveFitness(fitness);                       % 存储当前群体平均适应度
    if mod(iter, WaitbarInter) == 0                                             % 处理并显示进度条
        waitbar(iter / IterMax, h, ['已完成:' num2str(iter / IterMax * 100) ... 
        '%   运算中...用时:', num2str(toc),'/',num2str(toc/(iter / IterMax))])
    end
end
close(h)
disp(["搜索完成，开始打印结果······"])

[ErgodicBestIndividual,ErgodicBestFitness] = PlotResult(ChromNum,IterMax,AveFitnessProgress,FitnessBestProgress);   % 绘制GA搜索结果和遍历结果
PrintResult(IndividualBest,FitnessBest,ErgodicBestIndividual,ErgodicBestFitness);                                   % 打印GA搜索结果和遍历结果

disp('打印完成！！！')







% 适应度计算
function fitness = CalcFitness(Individual)
    fitness = 2*cos(Individual(:,1).*Individual(:,2))+4*Individual(:,1)+Individual(:,2).^2;
end

% 初始化群体
function Individual = InitIndividual(N,ChromNum,DownLimit,UpLimit)
    Individual = zeros(N,ChromNum);
    for i = 1:ChromNum
        Individual(:,i) = rand(N,1)*(UpLimit(i)-DownLimit(i))+DownLimit(i);
    end
end

% 寻找当前群体最优个体
function [IndividualBest,FitnessBest] = FindBest(Individual,fitness)
    [FitnessBest,minCor] = min(fitness);
    IndividualBest = Individual(minCor,:);
end   

% 变异操作
function New_Individual = MutIndividual(Individual,iter,IterMax,Pm,DownLimit,UpLimit)
    [N,ChromNum] = size(Individual);
    % 遍历群体中的每一个个体
    for i = 1:N
        for j = 1:ChromNum
            if rand <= Pm               % 依照变异概率变异（可以实现自适应的变异概率，使该概率随着迭代次数越大变得越小，查文献找公式）
                Mut_num = rand * (1-iter/IterMax)^2; % 概率性的变异数值
                if rand<=0.5            % 按照50%概率决定变异方向
                    Individual(i, j) = Individual(i, j) * (1 - Mut_num);
                else
                    Individual(i, j) = Individual(i, j) * (1 + Mut_num);
                end
            end
        end
    end
    New_Individual = BoundaryLimit(Individual,DownLimit,UpLimit);
end

% 交叉操作
function New_Individual = CroIndividual(Individual,Pc)
    [N,ChromNum] = size(Individual);
    % 遍历群体中的每一个个体
    for i = 1:N
        if rand < Pc                % 依照交叉概率交叉（可以实现自适应的交叉概率，使该概率随着迭代次数越大变得越小，查文献找公式）
            Cro_Individual = floor((N-1)*rand+1);  % 被交叉的个体
            Cro_Node = floor((ChromNum-1)*rand+1); % 被交叉的染色体节点
            % 交叉操作
            temp = Individual(Cro_Individual,Cro_Node);
            Individual(Cro_Individual,Cro_Node) = Individual(i,Cro_Node);
            Individual(i,Cro_Node) = temp;
        end
    end
    New_Individual = Individual;
end

% 自然选择
function [New_Individual,New_Fitness] = NaturalSlection(Individual,fitness,IndividualBest,FitnessBest)
    [N,~] = size(Individual);
    SumFitness = sum(fitness);
    FitnessRate = fitness / SumFitness;
    for i = 1:N
        % 赌轮盘形式选择
        if rand >= FitnessRate(i)
            % 被选择的个体用概率当前最优个体替换
            fitness(i) = FitnessBest;
            Individual(i,:) = IndividualBest;
        end
    end
    New_Individual = Individual;
    New_Fitness = fitness;
end 

% 计算群体平均适应度
function AveFitness = CalcAveFitness(fitness)
    [N,~] = size(fitness);
    AveFitness = sum(fitness) / N; 
end

% 边界条件
function New_Individual = BoundaryLimit(Individual,DownLimit,UpLimit)
    [N,ChromNum] = size(Individual);
    for i = 1:N
        for j = 1:ChromNum
            if Individual(i,j) > UpLimit(j)
                Individual(i,j) = UpLimit(j);
            end
            if Individual(i,j) < DownLimit(j)
                Individual(i,j) = DownLimit(j);
            end
        end
    end
    New_Individual = Individual;
end

% 绘制GA搜索结果和遍历结果
function [ErgodicBestIndividual,ErgodicBestFitness] = PlotResult(ChromNum,IterMax,AveFitnessProgress,FitnessBestProgress)
    ErgodicBestIndividual = zeros(1,ChromNum);
    figure(1)
    plot(1:IterMax,AveFitnessProgress,'r',1:IterMax,FitnessBestProgress,'b');
    grid on
    legend('平均适应度','最优适应度')
    x = -5:0.01:5;
    y = -5:0.01:5;
    z = 2*cos(x.*y)+4*x+y.^2;
    figure(2)
    scatter3(x,y,z,'ko');
    [X,Y] = meshgrid(x, y);
    Z = 2*cos(X.*Y)+4*X+Y.^2;
    mesh(X,Y,Z);
    [a,b]=find(Z==min(min(Z)));
    ErgodicBestFitness = Z(a(end),b(end));
    ErgodicBestIndividual(1,1) = X(a(end),b(end));
    ErgodicBestIndividual(1,2) = Y(a(end),b(end));
end

% 打印GA搜索结果和遍历结果
function PrintResult(IndividualBest,FitnessBest,ErgodicBestIndividual,ErgodicBestFitness)
    disp(['GA搜索结果：']);
    disp(['     自变量：x1 = ',num2str(IndividualBest(1)),'  , x2 = ',num2str(IndividualBest(2))]);
    disp(['     最小值：f = ',num2str(FitnessBest)]);
    disp(['Ergodic搜索结果：']);
    disp(['     自变量：x1 = ',num2str(ErgodicBestIndividual(1)),'  , x2 = ',num2str(ErgodicBestIndividual(2))]);
    disp(['     最小值：f = ',num2str(ErgodicBestFitness)]);
end