ceclpso.m

code for combined effect comprehensive learning PSO

function out=ceclpso(functchc,D,Maxgen,epochtime,Rmax,Rmin,Stop_criteria)
% Combined Effect Comprehensive learning PSO
global initial_flag EpochtimeArr tr FE_success_runs success_runs M O
global result_type
% Parameter Settings
NP=25;Xmax(1:D)=Rmax;Xmin(1:D)=Rmin;
Vmax = 0.25*(Xmax - Xmin);
P=0.07;m=0.20*D; wmax=0.9; wmin=0.2;w0=0.9; w1=0.2;
%------------------------------

%--- Inertia weight And C1,C2,C3 decreases linearly ----------------
for iter=1:Maxgen
    W(iter)=( (w0-w1)*(Maxgen -iter))/Maxgen + w1;
end
    

%---------- Initialization of swarm -------------
for count_p = 1:NP
    for count_d= 1:D
        X(count_p,count_d) = rand;
        V(count_p,count_d) = rand;
        p_best(count_p,count_d) = X(count_p,count_d);
    end
end
%------------------------------
%-------------Evaluate fitness value pbest and gbest of all particles
for count_p = 1:NP
            for count_d= 1:D
               xy(count_d)=X(count_p,count_d);
            end
        current_fitness(count_p)=BENCH_FUNCS(xy,functchc);
end 
                      
    [g_best_val,g_best_index] = min(current_fitness);
    for count_d = 1:D
      g_best(count_d) = X(g_best_index,count_d);
    end
%-----------------------------------------------
%-------   Second gbest ------------
 temp=current_fitness;
                 temp(g_best_index)=max(current_fitness);
    [g_best_val_2,g_best_index_2] = min(temp);                 
                 for count_d= 1:D
                       g_best_2(count_d) = X(g_best_index,count_d);
                 end
%-------------------------------------

if epochtime ==1
    tic;
end
%-------- Main routine starts ---------------
for count_g = 1:Maxgen
    %------------------------------
        if mod(count_g,10) == 1
                for count_p=1:NP
                    rc=randperm(D);
                    a(count_p,:)=zeros(1,D);
                    b(count_p,:)=zeros(1,D) ;      
                    a(count_p,rc(1:m))=1;
                    b(count_p,:)=ceil(rand(1,D)-1+P);
                    f(count_p,:)=ceil(rand(1,D).*NP);
                end
        end
      %-------------------------------
       %----------------------------- Finding the Trapped Particle 
        if mod(count_g,10) == 1
                     if count_g~=1
                            sorted_fitness=sort(current_fitness);
                             k=floor(NP./2);
                             N=k;
                                   if mod(NP,2)~=0
                                        k=k+1;
                                   end                                  
                         for m=1:N
                             for q=1:NP
                                    dif =sorted_fitness(k+m)-current_fitness(q);
                                       if dif==0
                                           for count_d=1:D
                                              X(q,count_d)=rand*g_best(count_d);
                                           end
                                       end
                             end
                         end
                     end
        end
                                             
 %----------------------------------------------------------------       
   
    % ----------  Velocity and Possition Updation
    for count_p = 1:NP
        for count_d= 1:D  
               if a(count_p,count_d)==1
V(count_p,count_d)=W(count_g)*V(count_p,count_d)+rand*(g_best(count_d)-X(count_p,count_d))+rand*(g_best_2(count_d)-X(count_p,count_d));
                  else if b(count_p,count_d)==1
V(count_p,count_d)=W(count_g)*V(count_p,count_d)+rand*(p_best(f(count_d),count_d)-X(count_p,count_d))+rand*(g_best_2(count_d)-X(count_p,count_d));
                       else
V(count_p,count_d)=W(count_g)*V(count_p,count_d)+rand*(p_best(count_p,count_d)-X(count_p,count_d))+rand*(g_best_2(count_d)-X(count_p,count_d));
                      end
               end
                                       
  V(count_p,count_d)=min(Vmax(count_d),max(-Vmax(count_d),V(count_p,count_d)));
                     X(count_p,count_d)=X(count_p,count_d)+V(count_p,count_d);
        end
    end      
%------------------------------------------------------------------------
             
 % -------- Computing f(xi) ---------
    for count_p = 1:NP          
                for count_d= 1:D
                     xy(count_d)=X(count_p,count_d);
                end
              current_fitness(count_p)=BENCH_FUNCS(xy,functchc);
    end   
% -------- updating pbest---------    
    for count_p = 1:NP
        for count_d= 1:D
          xy(count_d)=p_best(count_p,count_d);
        end
       p_best_fitness(count_p)=BENCH_FUNCS(xy,functchc);        
        
            if current_fitness(count_p)< p_best_fitness(count_p)
                     p_best_fitness(count_p) = current_fitness(count_p);
                        for count_d = 1:D
                            p_best(count_p,count_d) = X(count_p,count_d);
                        end
           end
    end
%-----------------------------
%--------- Updating gbest------------
      for count_d= 1:D
          xy(count_d)=g_best(count_d);
      end
       g_best_fitness(count_g)=BENCH_FUNCS(xy,functchc);        
 
       [g_best_val,g_best_index] = min(p_best_fitness);
       if g_best_val < g_best_fitness(count_g)
           g_best_fitness(count_g)=g_best_val;
           for count_d = 1:D
                   g_best(count_d) = p_best(g_best_index,count_d);
           end        
       end     
%---------------------------------------                     
%-------   Updating Second gbest ------------
    for count_d= 1:D
          xy(count_d)=g_best_2(count_d);
    end    
       g_best_fitness_2(count_g)=BENCH_FUNCS(xy,functchc);        
  temp=p_best_fitness;
                  temp(g_best_index)=max(p_best_fitness);
     [g_best_val_2,g_best_index_2] = min(temp);   

        if g_best_val_2 < g_best_fitness_2(count_g)
           g_best_fitness_2(count_g)=g_best_val_2;
           for count_d = 1:D
                   g_best_2(count_d) = p_best(g_best_index_2,count_d);
           end        
       end          

            % -------- Stop criteria ---------
             if g_best_fitness(count_g) <Stop_criteria   
             disp('Stop criteria');
             success_runs= success_runs+1;
             FE_success_runs(success_runs)=g_best_fitness(count_g);
                break
             end
           %---------------------------------
                    
end
if epochtime ==1
    EpochtimeArr(tr)=toc;
end

for n=count_g:Maxgen
    g_best_fitness(n)=g_best_fitness(count_g);
end

Value=g_best;
                
        if result_type==1
            out=g_best_fitness;
            FitnessCECLPSO=g_best_fitness(Maxgen)
        else
            out=g_best_fitness(Maxgen);  
        end