Fmincon seems to repeat a number of times the same set of variables

Hello,

Fmincon seems to repeat a number of times the same set of variables.

Not sure why?

What am I missing?

Thanks

Answer 1

[Steven Lord]

Do you mean that fmincon appears to evaluate your function at the same point multiple times?

If so, how have you checked that those points are actually identical? It's possible they are very slightly different, but that the change is small enough that it doesn't change how the number is displayed in the default format if that's how you're checking. In the example below a and b are displayed the same but their stored values are not the same.

a = pi

a = 3.1416

b = a + 1e-6

b = 3.1416

b == a % false

ans = logical
   0

[Robert Jones]

I see. I have displayed only 2 digits. I will extend to 6 digits

Thank you

[Torsten]

@Robert Jones

In a fixed point x0, fmincon has to calculate the gradient of your objective function and the Hessian numerically. This is done via finite-difference approximations:

df/dxi @ x=x0 ~ (f(x10,...,xi0+h,...,xn0) - f(x10,...,xi0,...,xn0))/h

So if you output x in your objective function with a limited number of digits, you might get the impression that

(x10,...,xi0+h,...,xn0) equals (x10,...,xi0,...,xn0).