Fitness function run twice / stopping a GA

Hello.

I have currently two problem with my GA code. My elements are defined by 4 value, and each time my fitness function is called, I display the values. What I get is :

   -0.1827    0.2077   -0.0817    0.0275
   -0.1827    0.2077   -0.0817    0.0275
    0.8208   -0.6566    0.1232    0.0194
   -0.8640    0.7799   -0.1599    0.0242

As you can see, the first element is plotted twice,meaning the fitness function is called twice with that element. But if I look at the population, it is not in it.

Here is my "run" code :

 clear all
ObjectiveFunction = @fitnescalc;
nvars = 4;    % Number of variables
ConstraintFunction = @constraint;
CreationFcn=@populationInit;
IP=[populationInit()]
for i=1:2
IP = [IP;populationInit()]; % Population MAtrix goes here
end
options = optimoptions('ga','InitialPopulationMatrix',IP,'OutputFcn',@gaoutfun,'PopulationSize',length(IP(:,1)));
[x,fval] = ga(ObjectiveFunction,nvars,[],[],[],[],[],[], ...
    ConstraintFunction,options);

Do you see where it could come from ? Thank you

[Geoff Hayes]

Boris - what does populationInit() do? i.e. what is the code for this function? Does it return a single element or multiple ones? It looks like you initialize IP as

 IP=[populationInit()]

and then update it twice with two calls to this same function so how large is your initial population?

Please copy and paste the code for populationInit to your question.

[Boris Huljak]

he code is the following :

 function [ p_shape ] = populationInit()
 p_shape_degree=3;
 %%%The computed polynomial degree will be : p_shape_degree + 1
 % Constants declaration
 Xlenght=60e-2;
 Ylenght=9e-2; %Diametre TOTAL
 X_max = Xlenght;
 Y_min = 1e-2; % Minimum radius accepted
 Y_max = Ylenght/2; % Maximum radius accepted
 thickness_min =10e-3; % Minimum thickness accepted
is_poly_ok = 0;
iteration_nb = 1; % Convergence security
 X = 0:.05:X_max;
 % polynomial computation
 while(~is_poly_ok)
Y_in = Y_min + rand()*(Y_max - Y_min);
  Y_out = Y_min + rand()*(Y_max - Y_min);
    p_in = [-1/X_max 1];
    p_out = [1/X_max 0];
    p = Y_in*p_in + Y_out*p_out;
    p_fit = zeros(1, p_shape_degree + 1);
    p_fit(p_shape_degree : p_shape_degree + 1) = p;
    % Shape polynomial
    p_conv = 5*randn(1, p_shape_degree - 1)/iteration_nb;
    for i = 1:p_shape_degree-1
        p_conv(i) = p_conv(i)/(X_max^(p_shape_degree - 1 - i));
    end
    p_shape =  p_fit + conv(conv(p_in, p_out), p_conv);
    vals = polyval(p_shape, X);
    if (min(vals)> Y_min && max(vals) < Y_max - thickness_min)
        is_poly_ok = 1;
    end
    iteration_nb = iteration_nb + 1;
end
end

It is a bit messy, but what it does, it simply returns a vector of the form p=[A B C D].

My initial population is weirdly made, I agree, but once I inspect the vector, it seems fine with no duplicate