Re: [Help-glpk] [Fwd: Tricks with MathProg to approximate non-linear functions?]

[Chris Matrakidis]

Dear Dolan,

You mail had to be forwarded manually because you are not subscribed
to the list.

> Are there any known tricks with GLPK for what I'm trying to do, or am I
> best off just choosing a linear objective function?

You don't give enough information for detailed answers, so some
general comments:
1. For piecewise linear approximations, you can use SOS2 constraints,
which you can model using the approach in
http://winglpk.sourceforge.net/media/glpk-sos2_02.pdf . Note that if
you have a linear program this will make it an integer one.
2. The FICO guide you reference shows several linearisations that you
can use if you have binary variables.
3. For a single ratio, like in the F1 score, the method described in
http://lpsolve.sourceforge.net/5.5/ratio.htm may be appropriate.
4. You can evaluate the nonlinear function after the optimisation and
compare it to the approximate one to see how accurate it was.

Hope this helps.

Best Regards,

Chris Matrakidis



>
> Best Regards,
> Dolan Antenucci
>
>
>
>
>
>
>
> _______________________________________________
> Help-glpk mailing list
> address@hidden
> https://lists.gnu.org/mailman/listinfo/help-glpk