sga_pro.m

一个非常好的遗传算法参考程序,可以优化带约束条件的函数.

function [maxfitness,minfitness,meanfitness,best_decimal_space,now_generation,now_probability_crossover,best_binary_space]=SGA_pro(min_confines,max_confines,probability_crossover,probability_mutation,population,decimal_step,max_generation,convergence_method,max_no_change_probability_crossover_generation,deta_fitness_max,deta_fitness_max_min,max_probability_crossover,max_no_change_fitness_generation,probability_crossover_step)


%SGA Function Of Simple Genetic Algorithm Program (Version 1.0.0.1)
%support multi-parameters
%By chen yi ,CQU .QQ:2376635  Email:[e-mail]  (Aug,9th,2002)
%~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
%SGA function is the main function of  Simple Genetic Algorithm Program
%usage:
%[maxfitness,minfitness,meanfitness,best_decimal_space,now_generation]=SGA_pro(min_confines,max_confines,probability_crossover,probability_mutation,population,decimal_step,max_generation)
%
%min_confines is the minimum of input value in decimal-space
%max_confines is the maximum of input value in decimal-space
%i.e. X belong to [min_confines,max_confines]
% min_confines < max_confines
%~~
%decimal_step is the search step in decimal-space
%it must be smaller than (max_confines-min_confines)/2
%~~~
%population is the random decimal value in  [min_confines,max_confines]
% it is given by yourself and it must be Even integer number and larger than 0
%~~~~
%probability_crossover is the crossover probability in crossover step,
%your can give it's value to reach your need (About:0.4~0.99)
%
%probability_mutation is given by yourself (about:0.001~0.01)

%  See Also DECODING ,CODING ,CROSSOVER,MUTATION,FITNESS,
%           FITNESS_FUNCTION, 	SELECTION	 

%~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
%
[binary_space,bits_sum,bits]=coding(min_confines,max_confines,population,decimal_step);

[decimal_space]=decoding(min_confines,max_confines,binary_space,bits);

%[population,parameter_numbers]=size(decimal_space);

%ini
Generation_numbers=1;
no_change_fitness_generation=1;
no_change_probability_crossover_generation =1;
now_generation=max_generation;
fitness_plot_max=zeros(1,now_generation);
fitness_plot_min=zeros(1,now_generation);
fitness_plot_mean=zeros(1,now_generation);

%best_decimal_space=zeros(population,parameter_numbers);
%ini over

if convergence_method==1
   disp('convergence_method 1 is Max generation !');
   while Generation_numbers<=max_generation
%%%%%%%%%%%%% Co-evolutionary Genetic Algorithms 
[subjected_decimal_value]=Subject_fitness(decimal_space);
[fitness_value]=fitness(subjected_decimal_value);
%%%%%%%%%%%%%%% end Co-GAs

fitness_plot_max(1,Generation_numbers)=max(fitness_value);
fitness_plot_min(1,Generation_numbers)=min(fitness_value);
fitness_plot_mean(1,Generation_numbers)=mean(fitness_value);
[max_fitness_temp_position,decimal_space,binary_space,maxfitness]=selection(decimal_space,binary_space,fitness_value,bits);
[decimal_space,binary_space]=crossover(min_confines,max_confines,decimal_space,binary_space,bits,probability_crossover);
% pay attention to the decimal_space and binary_space from mutation.m 
% it can't be  [decimal_space_mutation,binary_space_mutation]
[decimal_space,binary_space]=mutation(min_confines,max_confines,decimal_space,binary_space,bits,probability_mutation);
Generation_numbers=(Generation_numbers+1);
best_binary_space=binary_space;
end % end of while 

best_decimal_space=decimal_space(max_fitness_temp_position(population),:);
maxfitness1=fitness(best_decimal_space);
maxfitness2=max(fitness_plot_max);

if maxfitness1==maxfitness2
    maxfitness=maxfitness1;
end

minfitness=min(fitness_plot_min);
meanfitness=mean(fitness_plot_mean);
now_generation=Generation_numbers-1;
now_probability_crossover=probability_crossover;

%plot
hold on;
ii=1:1:max_generation;
fitness_plot=plot(ii,fitness_plot_max,'k',ii,fitness_plot_min,'r',ii,fitness_plot_mean,'b');
grid on
title('Every Fitness--population')
xlabel('population');
ylabel('every-fitness');
legend('max','min','mean',4);

% end of convergence 1

% begin convergence 2

elseif convergence_method==2
disp('convergence_method 2 is no change fitness !')
 while (no_change_probability_crossover_generation <=max_no_change_probability_crossover_generation)
%%%%%%%%%%%%% Co-evolutionary Genetic Algorithms 
[subjected_decimal_value]=Subject_fitness(decimal_space);   
[fitness_value]=fitness(subjected_decimal_value);
%%%%%%%%%%%%%%% end Co-GAs
fitness_plot_max(1,Generation_numbers)=max(fitness_value);
fitness_plot_min(1,Generation_numbers)=min(fitness_value);
fitness_plot_mean(1,Generation_numbers)=mean(fitness_value);

[max_fitness_temp_position,decimal_space,binary_space,maxfitness]=selection(decimal_space,binary_space,fitness_value,bits);
[decimal_space,binary_space]=crossover(min_confines,max_confines,decimal_space,binary_space,bits,probability_crossover);
% pay attention to the decimal_space and binary_space from mutation.m 
% it can't be  [decimal_space_mutation,binary_space_mutation]
[decimal_space,binary_space]=mutation(min_confines,max_confines,decimal_space,binary_space,bits,probability_mutation);

if Generation_numbers~=1
  if (abs(fitness_plot_max(1,Generation_numbers)-fitness_plot_max(1,Generation_numbers-1))~=deta_fitness_max)&...
      (abs(fitness_plot_max(1,Generation_numbers)-fitness_plot_min(1,Generation_numbers))~=deta_fitness_max_min)&...
      (abs(fitness_plot_min(1,Generation_numbers)-fitness_plot_min(1,Generation_numbers-1))~=deta_fitness_max)&...
      (abs(fitness_plot_mean(1,Generation_numbers)-fitness_plot_mean(1,Generation_numbers-1))~=deta_fitness_max)
        if no_change_fitness_generation>=max_no_change_fitness_generation
        if probability_crossover<=max_probability_crossover
       probability_crossover=probability_crossover+probability_crossover_step;
       no_change_probability_crossover_generation=no_change_probability_crossover_generation+1;
       end      
   end
no_change_fitness_generation=no_change_fitness_generation+1;
end
end
Generation_numbers=(Generation_numbers+1);
best_binary_space=binary_space;
end % end of while

best_decimal_space=decimal_space(max_fitness_temp_position(population),:);
maxfitness1=fitness(best_decimal_space);
maxfitness2=max(fitness_plot_max);

if maxfitness1==maxfitness2
    maxfitness=maxfitness1;
end
minfitness=min(fitness_plot_min);
meanfitness=mean(fitness_plot_mean);
now_generation=Generation_numbers-1;
now_probability_crossover=probability_crossover; 

else % other convergence method  update choice
    disp('The convergence method has two way now, chose:1 or 2');
    break;
end

%plot
hold on;
if now_generation>max_generation
   i=1:1:now_generation;
else
    i=1:1:(max_generation);
end
fitness_plot=plot(i,fitness_plot_max,'k',i,fitness_plot_min,'r',i,fitness_plot_mean,'b');
grid on
title('Every Fitness--population')
xlabel('population');
ylabel('every-fitness');
legend('max','min','mean',4);

% end of sga_pro.m