Since ‘objective function’ implies curve fitting, try something like this —
x = 1:0.1:10;
y = 2.5*exp(-(x-5).^2/2)+randn(size(x))*0.25;
objfcn = @(b,x) b(1).*exp(-(x-b(2)).^2*b(3));
fitnessfcn = @(b) norm(y-objfcn(b,x));
Parms = 3;
[B,fval] = ga(fitnessfcn, Parms)
figure
plot(x, y, '.')
hold on
plot(x, objfcn(B,x), '-r')
hold off
grid
The parameter estimates here (2.69, 4.97, 0.666) are reasonably accurate when compared to the actual parameters (2.5, 5.0, 0.5) in this relatively simple problem. The norm of the residuals is 2.71.
.
