Fval Output by ga does not match output of fitness function evaluated at the respective x vector

Hello all,

I have been using ga to minimize a fitness function of mine, but have recently come across some issues concerning the output. After running for several hours (the fitness function is quite dense), the ga outputs an x vector along with an fval, which is supposed to be the fitness function evaluated at this x value. However, when I take the x-vector and input it into the fitness function manually, it returns a value different from the fval.

My ga setup is the following:

%Contraints and function
rng default
FitFcn = @FullDevice;
nvars = 3;
lb = [1,20,5];
ub = [5,100,10];
IntCon = [1]; 

%optomization algorithm
[x,fval,exitflag] = ga(FitFcn,nvars,[],[],[],[],lb,ub,[],IntCon,[])

As an example, on my most recent optimization, the x and fval output were the following

x =

    3.0000   20.0000    5.0333


fval =

 -199.1417

But,

 >  FullDevice([3 20 5.0333])

ans =

 -189.8933

Any thoughts as to why this is happening? Thanks!

Answer 1

[Stephan]

you problably are are victim of rounding errors. try:

FullDevice(x)

to use the exact results of x as input for your Fitness function.

[Francisco Barrera]

This was it. Silly oversight on my part. Thanks for the help!