[Matlab] 二进制蝙蝠算法用于解决背包问题

算法流程

背包问题

参考[Matlab] 离散二进制粒子群算法（BPSO ）解决背包问题

Matlab 代码

%% 二进制蝙蝠算法

%初始化种群
r = 0.7;
narvs = 10; %问题的维度
n = 100;
dim = narvs;
maxgen = 100;
populationSize = 100;

Qmin=0;         % Frequency minimum 频率最小值
Qmax=2;         % Frequency maximum 频率最大值

Q=zeros(n,1);   % Frequency
v=zeros(n,dim);   % Velocities 速度

cg_curve=zeros(1,maxgen);

x = randsrc(n,narvs,[0,1;0.5,0.5]);

%计算适应度值
fitness =targetPackage(x',100);

%找到当前最好
[fmax,I]=max(fitness);
best=x(I,:);

A = 0.6;
bestfit = zeros(maxgen,1);
for i = 1:maxgen
    cg_curve(i)=fmax; %当前值
    for j = i:populationSize
        for d = 1:dim
            Q(i) = Qmin+(Qmin-Qmax)*rand; %更新频率
            v(j,d) = v(j,d)+(x(j,d)-best(d))*Q(j); %更新速度
            % V型的转换函数，很多采用的是西格玛函数用于速度到位置的离散转换
            V_shaped_transfer_function=abs((2/pi)*atan((pi/2)*v(j,d))); 
            if rand<V_shaped_transfer_function % Equation 10 in the paper
                x(j,d)=~x(j,d);
            else
                x(j,d)=x(j,d);
            end
            if rand > r  % Pulse rate ()
                x(j,d)=best(d);
            end
        end
        Fnew = targetPackage(x(j,:)',1);
        if (Fnew >= fitness(j,:))&&(rand < A)
            x(j,:)=x(j,:);
            fitness(j,:) = Fnew;
        end
        if Fnew>=fmax
            best=x(j,:);
            fmax=Fnew;
        end
    end 
    bestfit(i,:) = fmax;
end

targetPackage(best',1)
plot(bestfit)

目标函数

function fitness = targetPackage(x,indNum)
%targetPackage 此处显示有关此函数的摘要
%   X：表示种群
volume=[95 75 23 73 50 22 6 57 89 98];        %物品体积
value=[89 59 19 43 100 72 44 16 7 64];          %物品价值
Weight=300;  %背包重量
% a = zeros(indNum,1);
% for i=1:indNum
%     a(i,1) = volume*x(:,i);
% end
%超过总重量的个体的适应度值都视为0
k = find(volume*x<Weight);
fitness=zeros(indNum,1);
for j=1:size(k,2)
    fitness(k(j),1) = value*x(:,k(j));
end
end

仿真结果

参考文献

[1] Mirjalili, Seyedali, Seyed Mohammad Mirjalili, and Xin-She Yang. 2014. “Binary Bat Algorithm.” Neural Computing and Applications 25 (3): 663–81.