0

I have a blackbox optimization problem, meaning, I have a shared-object file that I can call from within a Matlab function supplying it with input variable x and I can get the objective value obj as well as constraint-violations cv. The following can be noted :

  1. x is a vector of known length with known bounds
  2. obj is a scalar value
  3. cv is a vector of known length

There is no information available on type of constraints. Can use fmincon to solve this optimization problem.

P.S : I already have a good starting point.

Abhinav
  • 1,882
  • 1
  • 18
  • 34
  • I think you can. However note that `fmincon` will do finite differences. `fmincon` really assumes your functions are smooth, exact and cheap to evaluate. If your evaluations are expensive you may want to use a derivative free method instead. – Erwin Kalvelagen Sep 27 '16 at 00:59
  • @ErwinKalvelagen Actually, my shared-lib function also returns gradients. Can I use `nonlcon` option? – Abhinav Sep 27 '16 at 01:11
  • Exact gradients will help `fmincon` a lot. – Erwin Kalvelagen Sep 27 '16 at 01:58

0 Answers0