Optimization: which function is better for my case?

I have to perform an optimization with 3 inputs over a cost function. The 3 inputs have precise boundaries and stay around 1. I have tried:

- fminbnd, but it seems to not accept vector x;

- fmincon, but it shows me an error (see figure below);

- fminsearch, which has a good performance, but of course the result doesn't satifies every time the constraint.

Do you have suggestions?

Answer 1

[Kirby Fears]

From the fmincon screen shot, you didn't get any errors - just a warning. It's more of a notification.

fmincon is the way to go. Go through the documentation carefully to insure that you're supplying inputs as intended. Keep an eye on setting the tolerances and boundaries appropriately.

Answer 2

[Abdelmoumen Bacetti]

It would be better if you can provide some code.

From the output, and regarding the narrow interval of your variables, I suspect that your cost function is constant or the step size is not small enough to allow the algorithm find a decreasing direction.

If you had chosen a small step, make sure that your function is convex in the interval by plotting it.