%% 清空环境变量
clear
clc
close all
warning off
%% 导入数据
load data
data=[attributes’ strength’];
% 输入数据
input =data(:,1:4)‘;
output=data(:,5)’;
% 随机生成训练集、测试集
k = randperm(size(input,2));
N = 24;
% 训练集
P_train=input(:,k(1:N));
T_train=output(k(1:N));
% 测试集
P_test=input(:,k(N+1:end));
T_test=output(k(N+1:end));%% 归一化
% 训练集
[Pn_train,inputps] = mapminmax(P_train);
Pn_test = mapminmax(‘apply’,P_test,inputps);
% 测试集
[Tn_train,outputps] = mapminmax(T_train);
Tn_test = mapminmax(‘apply’,T_test,outputps);%% 参数设置
popsize = 200; % 种群大小
bestsize = 5; % 优胜子种群个数
tempsize = 5; % 临时子种群个数
SG = popsize / (bestsize+tempsize); % 子群体大小
S1 = size(Pn_train,1); % 输入层神经元个数
S2 = 5; % 隐含层神经元个数
S3 = size(Tn_train,1); % 输出层神经元个数
iter = 10; % 迭代次数%% 随机产生初始种群
initpop = initpop_generate(popsize,S1,S2,S3,Pn_train,Tn_train);%% 产生优胜子群体和临时子群体
% 得分排序
[sort_val,index_val] = sort(initpop(:,end),‘descend’);
% 产生优胜子种群和临时子种群的中心
bestcenter = initpop(index_val(1:bestsize)😅;
tempcenter = initpop(index_val(bestsize+1:bestsize+tempsize)😅;
% 产生优胜子种群
bestpop = cell(bestsize,1);
for i = 1:bestsize
center = bestcenter(i,:);
bestpop{i} = subpop_generate(center,SG,S1,S2,S3,Pn_train,Tn_train);
end
% 产生临时子种群
temppop = cell(tempsize,1);
for i = 1:tempsize
center = tempcenter(i,:);
temppop{i} = subpop_generate(center,SG,S1,S2,S3,Pn_train,Tn_train);
endwhile iter > 0
%% 优胜子群体趋同操作并计算各子群体得分
best_score = zeros(1,bestsize);
best_mature = cell(bestsize,1);
for i = 1:bestsize
best_mature{i} = bestpop{i}(1,:);
best_flag = 0; % 优胜子群体成熟标志(1表示成熟，0表示未成熟)
while best_flag == 0
% 判断优胜子群体是否成熟
[best_flag,best_index] = ismature(bestpop{i});
% 若优胜子群体尚未成熟，则以新的中心产生子种群
if best_flag == 0
best_newcenter = bestpop{i}(best_index,:);
best_mature{i} = [best_mature{i};best_newcenter];
bestpop{i} = subpop_generate(best_newcenter,SG,S1,S2,S3,Pn_train,Tn_train);
end
end
% 计算成熟优胜子群体的得分
best_score(i) = max(bestpop{i}(:,end));
end
% 绘图(优胜子群体趋同过程)
figure(1)
temp_x = 1:length(best_mature{1}(:,end))+5;
temp_y = [best_mature{1}(:,end);repmat(best_mature{1}(end),5,1)];
plot(temp_x,temp_y,‘b-o’)
hold on
temp_x = 1:length(best_mature{2}(:,end))+5;
temp_y = [best_mature{2}(:,end);repmat(best_mature{2}(end),5,1)];
plot(temp_x,temp_y,‘r-^’)
hold on
temp_x = 1:length(best_mature{3}(:,end))+5;
temp_y = [best_mature{3}(:,end);repmat(best_mature{3}(end),5,1)];
plot(temp_x,temp_y,‘k-s’)
hold on
temp_x = 1:length(best_mature{4}(:,end))+5;
temp_y = [best_mature{4}(:,end);repmat(best_mature{4}(end),5,1)];
plot(temp_x,temp_y,‘g-d’)
hold on
temp_x = 1:length(best_mature{5}(:,end))+5;
temp_y = [best_mature{5}(:,end);repmat(best_mature{5}(end),5,1)];
plot(temp_x,temp_y,‘m-*’)
legend(‘子种群1’,‘子种群2’,‘子种群3’,‘子种群4’,‘子种群5’)
xlim([1 10])
xlabel(‘趋同次数’)
ylabel(‘得分’)
title(‘优胜子种群趋同过程’)%% 临时子群体趋同操作并计算各子群体得分
temp_score = zeros(1,tempsize);
temp_mature = cell(tempsize,1);
for i = 1:tempsize
temp_mature{i} = temppop{i}(1,:);
temp_flag = 0; % 临时子群体成熟标志(1表示成熟，0表示未成熟)
while temp_flag == 0
% 判断临时子群体是否成熟
[temp_flag,temp_index] = ismature(temppop{i});
% 若临时子群体尚未成熟，则以新的中心产生子种群
if temp_flag == 0
temp_newcenter = temppop{i}(temp_index,:);
temp_mature{i} = [temp_mature{i};temp_newcenter];
temppop{i} = subpop_generate(temp_newcenter,SG,S1,S2,S3,Pn_train,Tn_train);
end
end
% 计算成熟临时子群体的得分
temp_score(i) = max(temppop{i}(:,end));
end
% 绘图(临时子群体趋同过程)
figure(2)
temp_x = 1:length(temp_mature{1}(:,end))+5;
temp_y = [temp_mature{1}(:,end);repmat(temp_mature{1}(end),5,1)];
plot(temp_x,temp_y,'b-o')
hold on
temp_x = 1:length(temp_mature{2}(:,end))+5;
temp_y = [temp_mature{2}(:,end);repmat(temp_mature{2}(end),5,1)];
plot(temp_x,temp_y,'r-^')
hold on
temp_x = 1:length(temp_mature{3}(:,end))+5;
temp_y = [temp_mature{3}(:,end);repmat(temp_mature{3}(end),5,1)];
plot(temp_x,temp_y,'k-s')
hold on
temp_x = 1:length(temp_mature{4}(:,end))+5;
temp_y = [temp_mature{4}(:,end);repmat(temp_mature{4}(end),5,1)];
plot(temp_x,temp_y,'g-d')
hold on
temp_x = 1:length(temp_mature{5}(:,end))+5;
temp_y = [temp_mature{5}(:,end);repmat(temp_mature{5}(end),5,1)];
plot(temp_x,temp_y,'m-*')
legend('子种群1','子种群2','子种群3','子种群4','子种群5')
xlim([1 10])
xlabel('趋同次数')
ylabel('得分')
title('临时子种群趋同过程')
%% 异化操作
[score_all,index] = sort([best_score temp_score],'descend');
% 寻找临时子群体得分高于优胜子群体的编号
rep_temp = index(find(index(1:bestsize) > bestsize)) - bestsize;
% 寻找优胜子群体得分低于临时子群体的编号
rep_best = index(find(index(bestsize+1:end) < bestsize+1) + bestsize);
% 若满足替换条件
if ~isempty(rep_temp)
% 得分高的临时子群体替换优胜子群体
for i = 1:length(rep_best)
bestpop{rep_best(i)} = temppop{rep_temp(i)};
end
% 补充临时子群体，以保证临时子群体的个数不变
for i = 1:length(rep_temp)
temppop{rep_temp(i)} = initpop_generate(SG,S1,S2,S3,Pn_train,Tn_train);
end
else
break;
end
%% 输出当前迭代获得的最佳个体及其得分
if index(1) < 6
best_individual = bestpop{index(1)}(1,:);
else
best_individual = temppop{index(1) - 5}(1,:);
end
iter = iter - 1;
