Development of an optimization function

Hello everyone,

I am working on the minimization of losses in the flow of the electrical network, In this context I will use the Firefly algorithm and I took the example below to understand its operating principle, but I get stuck on the line who asks me the fitness function (I tried with the optimization function Matlab ackleyfcn).

   function FFA 
  clear all
  clc
  N=input('Please Enter the Number of Firefly: ');
  clc
  maxGen=input('Please Provide the Stopping Criteria: ');
  clc
  alpha=input('Please Enter the value of alpha: ');
  clc
  Beta0=input('Please enter the Initial attractiveness: '); 
  clc 
  gamma=input('Please Provide the Absorption Coefficient (gamma): ');
  clc
  fitness=input('Select the test function: ');
  clc 
  FIREFLY=input('Please Enter the Population of the FIREFLY: or\nPress zero(0) for the default population: '); 
  clc
  if    FIREFLY==0
      pop=rand(N); 
  else 
      pop=FIREFLY;
  end
  popi=pop;
  popj=rand(N);
  Ii=objfunc(fitness,popi);
  %this reprsent the light intesity of(pop)
  %this reprsent the light intesity of(pop)
  Ij=objfunc(fitness,popj);
  %Ij=fitness+popj;
  itr=0; 
  while itr<maxGen 
      itr=itr+1; 
      for i=1:N 
          for j=1:N 
              r=sqrt((popi(i)-popi(j))^2+(popj(i)-pop(j))^2);
              if Ii<Ij 
                  beta=Beta0*exp(-gamma*r.^2);
                  Ii=popj(i).*(1-beta)+popj(j).*beta+alpha.*(rand-0.5);
              end
              beta=Beta0*exp(-gamma*r.^2);
              popi(i)=popi(i).*(1-beta)+popi(j).*beta+alpha.*(rand-0.5);
              popj(i)=popj(i).*(1-beta)+popj(j).*beta+alpha.*(rand-0.5); 
              Iinew=objfunc(fitness,popi(i)); Ijnew=objfunc(fitness,popj(i));
          end
      end
      if Iinew<Ijnew 
          Best_Soln=Iinew;
      else Best_Soln=Ijnew;
      end
  end
  disp(Best_Soln)

thank you in advance

[Walter Roberson]

Having "clear all" in code is rather like Wile E. Coyote exploding the bridge underneath himself and expecting to stay safe in mid-air as long as he does not look down.

[jelol sadf]

my question is about the fitness function

[Walter Roberson]

If you were to create a chip-level clone of your computer and a byte-level copy of your hard drive, then someone might be able to figure out exactly how your system would react to exploding the currently executing function. But short of that... Nope, not gonna happen. The byte fragments are going to land where-ever they land, and if they crash your MATLAB or mess up your program in some really strange way then it's nothing we can predict.

Answer 1

[Walter Roberson]

We do not know what objfunc() is, and what data types it expects so we cannot tell you what to put in for fitness . It might have to be a quoted string; or perhaps it should be an anonymous function. But we do not know how many arguments it should take.

It appears from the code that objfunc() expect to be able to be passed either individual scalar doubles or else square matrices.

rand(N) is an N x N matrix, so both popi and popj can be N x N matrices. But your nested for loop accesses popi(i) and popi(j) and popj(i) and popj(j) which is at most the first column of the N x N matrix. Either your for loop is incorrect or else your random initialization is wrong.

It would help if the input prompts gave some hints as to expected sizes.

Also,

              Iinew=objfunc(fitness,popi(i)); Ijnew=objfunc(fitness,popj(i));

Those overwrite all of Iinew and Ijnew in each iteration of the double nested for loops, and those variables are not otherwise used in the loop so they are not being constructed iterative. The end result is going to be whatever was set in the last iteration of the loops, when both i and j equal n. If that is what is desired then there is no point in calculating them inside the loops.

[jelol sadf]

Yes, I understand what you are saying. Do you have any proposals concerning fitness functions in the firefly algorithm? Are there preset optimization functions because from the function I can know the required inputs ?

[Walter Roberson]

objfunc = @(fun, x) fun(x);
fitness = @(x) (max(x) + min(x)) * coth(mean(x));   %a proposed fitness function