How to set up a complex objective function

[Julien Ruiz]

Hi,

I have an optimization problem (some kind of optimal mix) where I need to set weights for a set of real variables. I solved it with convex opt tool.

Then I would like to set up some constraints to find the optimal mix with a subset of the available variables (ie all to 0 except 5 of them).

I think I can solve it with chocosolver but I am stuck on how to set the objective function : I would like it to be a call to the convex optimization.

As far as I can sse it is not possible as the way to set the objective function is to have a closed form of it expressed as algebric manipulation of the other variables.

Is there a way around this ?

Brgds,

Julien

[cpru...@gmail.com]

Hi, 

Do you want some kind of nested resolution where the convex optimisation tool is used as a constraint ?

Best,

[Julien Ruiz]

Hi,

Thanks for the answer.

Yes it is a nested problem and - in the way I formulated - the convex optimization tool is used to define the objective function.

Something like :

Solve (on the xi) Min f(xi)

sc :

sum xi <n

xi in {0,1}

And f(xi) = Solve (on the ys) Min g(xi, ys)  -> computed by the optimization tools -> no closed form exists

I looked into bilevel ooptimization, but only found something for linear problems.

Brgds,

Julien

[cpru...@gmail.com]

Hi, 

In Choco, there will be no turnkey solutions to do this kind of operation. 

It will be necessary to create a system similar to the one set up with Ibex-lib, for example (which required development).

If you want to start on this kind of approach, we can probably guide you.

Best regards,

[Julien Ruiz]

Ok, thank you.

I'll check which is easiest for me : develop some cut-branch specific algo to my problem or try to integrate a callback in choco solver.

I'll keep you posted.