Hi Dale,
Thanks for your reply. Last night I added one additional constrains to make
the value of variables to be 0/1. Like this:
Subject: SUM{ (Xi * (Xi - 1))} = 0
I tried COBYLA, but it seems not working because it took a whole night
without finding a solution... (the dimensionality in my simulation is 76).
Do you have any other suggestions? Thanks again.
Best,
Kevin
2014-06-05 8:13 GMT-05:00 Smith, Dale (Norcross) <[email protected]>:
> This is a class of problems called mixed integer programming, although
> that term is most commonly used in the OR/linear programming field. I think
> the term we use here is nonlinear integer programming.
>
>
>
> http://en.wikipedia.org/wiki/Integer_programming
>
>
>
> http://arxiv.org/abs/0906.5171
>
>
>
> I didn’t find anything relevant by searching the list of algorithms in
> NLopt.
>
>
>
> http://ab-initio.mit.edu/wiki/index.php/NLopt_Algorithms
>
>
>
> It’s possible that your problem can be recast into something usable. I am
> thinking about a complementarity type problem for which a global solver
> with constraints can be used since you have a 0/1 variable.
>
>
>
> *Dale Smith, Ph.D.*
>
> Senior Financial Quantitative Analyst
>
> Financial & Risk Management Solutions
>
> Fiserv
>
> Office: 678-375-5315
>
> www.fiserv.com
>
>
>
> *From:* NLopt-discuss [mailto:[email protected]] *On
> Behalf Of *Bj Wu
> *Sent:* Thursday, June 05, 2014 1:08 AM
> *To:* [email protected]
> *Subject:* [NLopt-discuss] nonlinear optimization with dummy variables
>
>
>
> Hi guys,
>
>
>
> I am new to nlopt. I have a nonlinear optimization problem like:
>
>
>
> object: Max{ Sum((1-(1-Pi*x1)*(1-Pi*x2)*(1-Pi*x3))*Ri) } (0<i<4 and Pi, Ri
> is constant)
>
>
>
> subject: x1+x2+x3 <= 2
>
>
>
> xi is dummy variable which means it can only choose 0 or 1 value. Is there
> any API I can use to play with such dummy variables? For example, is it
> possible that I could use some API like the bounds setting APIs to fix the
> variables value to 0 or 1.
>
>
>
> Thank you very much.
>
>
>
> Best,
>
> Kevin
>
_______________________________________________
NLopt-discuss mailing list
[email protected]
http://ab-initio.mit.edu/cgi-bin/mailman/listinfo/nlopt-discuss
