1 简介

Biogeography-Based Optimizer (BBO) is employed as a trainer for Multi-Layer Perceptron (MLP). The current source codes are the demonstration of the BBO-MLP trainer for solving the Iris classification problem. There are also other trainers in this submission: Particle Swarm Optimization (PSO), Ant Colony Optimization (ACO), Genetic Algorithm (GA), Evolutionary Strategy (ES), and Probability-Based Incremental Learning (PBIL). The classification accuracy of BBO-MLP is calculated at the end of main.m file and compared to those of PSO, ACO, ES, GA, and PBIL. The convergence curve and classification accuracy of each algorithm are drawn at the end.

2 部分代码

%     Biogeography-Based Optimization (BBO) trainer for MLP       %
%                     source codes version 1                       %
%                                                                   %

clc
clear all

% For modifying initial parameters please have a look at init.m file 

display('..........................................................................................')
display('BBO is training MLP ...')
display('..........................................................................................')
[cg_curve1,Hamming,best] = BBO(@MLP_Iris, 1, 1, 300); % BBO trainer
Best_W_B(1,:)=best;

display('..........................................................................................')
display('PSO is training MLP ...')
display('..........................................................................................')
[cg_curve2,best]= PSO(@MLP_Iris, 1);  % PSO trainer
Best_W_B(2,:)=best;

display('..........................................................................................')
display('GA is training MLP ...')
display('..........................................................................................')
[cg_curve3,best]= GA(@MLP_Iris, 1); % GA trainer
Best_W_B(3,:)=best;

display('..........................................................................................')
display('ACO is training MLP ...')
display('..........................................................................................')
[cg_curve4,best]= ACO(@MLP_Iris, 1); % ACO trainer
Best_W_B(4,:)=best;

display('..........................................................................................')
display('ES is training MLP ...')
display('..........................................................................................')
[cg_curve5,best]= ES(@MLP_Iris, 1); % ES trainer
Best_W_B(5,:)=best;

display('..........................................................................................')
display('PBIL is training MLP ...')
display('..........................................................................................')
[cg_curve6,best]=PBIL(@MLP_Iris, 1); % PBIL trainer
Best_W_B(6,:)=best;

% Calculating classification rates
 
load iris.txt
x=sortrows(iris,2);

H2=x(1:150,1);
H3=x(1:150,2);
H4=x(1:150,3);
H5=x(1:150,4);
T=x(1:150,5);
 
H2=H2';
[xf,PS] = mapminmax(H2);  % Normalzation of input
I2(:,1)=xf;

H3=H3';
[xf,PS2] = mapminmax(H3); % Normalzation of input
I2(:,2)=xf;

H4=H4';
[xf,PS3] = mapminmax(H4); % Normalzation of input
I2(:,3)=xf;

H5=H5';
[xf,PS4] = mapminmax(H5); % Normalzation of input
I2(:,4)=xf;
Thelp=T;
T=T';
[yf,PS5]= mapminmax(T);   % Normalzation of output
T=yf;
T=T';

   for i=1:6
       Rrate=0;
       W=Best_W_B(i,1:63);
       B=Best_W_B(i,64:75);
       for pp=1:150
           actualvalue=my_MLP(4,9,3,W,B,I2(pp,1),I2(pp,2), I2(pp,3),I2(pp,4));
           if(T(pp)==-1)
               if (actualvalue(1)>=0.95 && actualvalue(2)<0.05 && actualvalue(3)<0.05)
                   Rrate=Rrate+1;
               end
           end
           if(T(pp)==0)
               if (actualvalue(1)<0.05 && actualvalue(2)>=0.95 && actualvalue(3)<0.05)
                   Rrate=Rrate+1;
               end  
           end
           if(T(pp)==1)
               if (actualvalue(1)<0.05 && actualvalue(2)<0.05 && actualvalue(3)>=0.95)
                   Rrate=Rrate+1;
               end              
           end
       end
       
       Final_Classification_Rates(1,i)=(Rrate/150)*100;
       
   end
   
display('--------------------------------------------------------------------------------------------')
display('Classification rate')
display('   BBO       PSO       GA       ACO       ES       PBIL')
display(Final_Classification_Rates(1:6))
display('--------------------------------------------------------------------------------------------')


figure('Position',[500 500 660 290])
%Draw convergence curves
subplot(1,2,1);
hold on
title('Convergence Curves')
semilogy(cg_curve1,'Color','r')
semilogy(cg_curve2,'Color','k')
semilogy(cg_curve3,'Color','b')
semilogy(cg_curve4,'Color','r')
semilogy(cg_curve5,'Color','g')
semilogy(cg_curve6,'Color','c')
xlabel('Generation');
ylabel('MSE');

axis tight
grid on
box on
legend('BBO','PSO', 'GA', 'ACO', 'ES', 'PBIL')

%Draw classification rates
subplot(1,2,2);
hold on
title('Classification Accuracies')
bar(Final_Classification_Rates)
xlabel('Algorithm');
ylabel('Classification rate (%)');

grid on
box on
set(gca,'XTickLabel',{'BBO','PSO', 'GA', 'ACO', 'ES', 'PBIL'});

3 仿真结果

4 参考文献

S. Mirjalili, S. M. Mirjalili, A. Lewis, Let A Biogeography-Based Optimizer Train Your Multi-Layer Perceptron, Information Sciences, In press, 2014