Can genetic algorithm be nested?

i want to have a code that does double optimization

i want the ga to first assume a vector q of 1s and 0s, then for each member in the population before calculating the fitness i want it to assume another vector which is the coefficients of q or if possible, only the elements that are assumed to be 1 in vector q. is a double optimization in the genetic algorithm possible ??

can i put a line that calls the ga toolbox in the fitness function for example ??

Answer 1

[Stephan]

It should be possible in one single call of ga. In this case nvars would be 2*numel(q) and you have to set the correct bounds and intcon correctly.

Then in the first half set of variables ga assumes 0/1 for q and the second half set is used for the assumed coefficients that are multiplied by the assumed q-values. Since some of them are 0 the multiplication with a coefficient would still be zero.

See this example:

function myfitness = myFitness(x)
q = x(1):x(numel(x)/2);
coeffs = x((numel(x)/2)+1):x(end);
q_coeffs = coeffs.*q;

% Caculate your fitness value
myfitness = q_coeffs * something - m*c^2; % Your fitness calculation...
end

What would happen for this example:

ga assumes:

x(1) = 1
x(2) = 0
x(3) = 1
x(4) = 0.1
x(5) = 0.5
x(6) = 0.25 
q_coeffs = [0.1*1,  0.5*0, 0.25*1] = [0.1 0 0.25]

[alaa zarif]

isnt there another way to assume coefficients for the values in q assumed to be 1 only ?? as like that i will have many unneccessary variables

in my problem i have a 64 variables to be assigned 0s and 1s, if i double them the code will consume so much time

[Stephan]

The problem is, that you need to know the number of vars before starting ga. But i doubt that it takes less time to start ga twice. 128 decision variables is not a big optimization problem. I would try this way, since it is easy to program and to understand. If you program it in a vectorized manner, this part should run quickly.