Using cost functions that aren't made of built-in matlab functions

[Stephanie]

Hi,

I would like to use YALMIP for a mixed-integer optimization problem on a linear hybrid piecewise affine dynamical system. The cost function that I want to use is nonlinear and nonconvex and does not have a closed form; it takes variables as input and runs through an iteration algorithm completely separate from YALMIP before returning a value. Is it possible to use such a cost function with YALMIP? If so, how? I don't see any examples of it on YALMIP's website.

Thanks!

Stephanie

[Johan Löfberg]

No, YALMIP is not designed for black-box models.

[Aashiq Muhamed]

Does it have this feature now? Or are the variables not capable of going through Monte Carlo or other iterative algorithms still?

[Johan Löfberg]

the only approach to use black.box functions is to use the sdpfun functionality, which is a hack and not efficient for anything but univariate functions