Matrix dimensions must agree

Hi,I have a problem with this program :

popsize=20; 

MAXITER=200; 

fevals=0;  

dimension=20; 

irange_l=-5.12; 

irange_r=5.12; 

xmax=100; 

xmin=0; 

M=(xmax-xmin)/2;

alpha=0.75; 

runno=2; 

f='my_Rastrigin';   

iter=0;

data=zeros(runno, MAXITER); 

   for run=1:runno

    x=(irange_r-irange_l)*rand(dimension,popsize,1)+ irange_l; 

    gbest=zeros(dimension,popsize); 

   for i=1:popsize

fpopul(i)=feval(f,x(:,i));

      fevals=fevals+1;
   end
   pbest=x;
   f_pbest=fpopul;
   [ f_gbest,g]=min(fpopul);
  while(iter<MAXITER)
      iter=iter+1;
      for n=1:MAXITER
     alpha=(1.0-0.5)*(MAXITER-n)/MAXITER+0.5; 
     mbest=sum(pbest)/popsize; 
     for i=1:popsize 
     fi=rand(1,dimension);
     p=fi*pbest(:,i)+(1-fi)*gbest; 
     u=rand(1,dimension);
     x(:,i)=p+((-1).^ceil(0.5+u)).*(alpha.*abs(mbest-x(:,i)).*log(1./u));
     x(:,i)=x(:,i)-(xmax+xmin)/2;
     x(:,i)=sign(x(:,i)).*min(abs(x(:,i)),M);
     x(:,i)=x(:,i)+(xmax+xmin)/2;
     fpopul(i)=feval(f,x(:,i)); 
     fevals=fevals+1;

     if (fpopul(i)<f_pbest(i))
     pbest(:,i)=x(:,i);
     f_pbest(i)=fpopul(i);
  end
     if f_pbest(i)<f_gbest
       gbest=pbest(:,i); 
       f_gbest=f_pbest(i);
     end
     end
  data(run,n)=f_gbest; 
      end
    [f_gbest,g]=min(f_pbest)

the peobleme is the dimension of matrix but I can not the corrected.

   x(:,i)=p+((-1).^ceil(0.5+u)).*(alpha.*abs(mbest-x(:,i)).*log(1./u));

please help me.

Answer 1

[Walter Roberson]

x(:,i) is a column.

mbest is sum(pbest)/popsize and pbest = x and x is two dimensional. sum() of a two dimensional array will give you a row vector, so mbest is going to be a row vector.

You have mbest-x(:,i) so that is a row vector minus a column vector. That is going to fail.

[mila boub]

Thank you for your answer, but how I can corrected this error? Because the formula of mbest is fixed.

[Walter Roberson]

If the formula for mbest is fixed then you need to rewrite everything else to work by row.

Have you considered the possibility of having mbest be the sum along rows?

Watch out by the way to be sure that u will match the shape of the result

[mila boub]

C formulas is fixed and the popul formula also

I try everything changed but I have nothing to do. this is the only error of the program.

for i=1:PopSize,

A(:,i)=bestpos(:,g);

end

R1=rand(dim,PopSize);

R2=rand(dim,PopSize);

C=sum(bestpos)/PopSize; 

alpha=(1.0-0.5)*(MaxIt-iter)/MaxIt+0.5; 

for i=1:PopSize,

fi=c1*R1/(c1*R1+c2*R2); 

p=fi*bestpos(i)+(1-fi)*fbestpart; 

u=rand(dim,PopSize);

   popul(:,i)=p+((-1).^ceil(0.5+fi)).*(alpha.*abs(popul(:,i)C).*log(1./u));

fpopul(i)=feval(f,popul(:,i));

      fevals=fevals+1;

end

the error matrix size is in this line:

popul(:,i)=p+((-1).^ceil(0.5+fi)).*(alpha.*abs(popul(i)-C).*log(1./u));

[Walter Roberson]

Please post your current code. The code you have posted so far is missing at least one "end", and the line

     popul(:,i)=p+((-1).^ceil(0.5+fi)).*(alpha.*abs(popul(:,i)C).*log(1./u));

is missing a "-" before the "C", indicating that you did not post your current code.