[SciPy-Dev] [SciPy-User] Linear Programming via Simplex Algorithm

[Christopher Jordan-Squire]

Aren't KNITRO and SNOPT specialized for nonlinear rather than linear problems?

On Sat, Nov 2, 2013 at 3:48 PM, Michael Yang <msyang at princeton.edu> wrote:
> Hi Rob & Alex - great job on the forward development of the simplex
> implementation.  I'm new to this thread but have been tracking its progress.
> Again, great job thus far and I'm looking forward to the final product.
>
> One recommendation for an alternate plan - and I realize this is coming late
> in the game and that you've made some excellent work - 'lp_solve' package
> does the simplex method as well as providing a full suite of linear
> programming features including the hard-to-implement integer constraints.
> It might be of interest to simply write a short python script that would
> convert the objective and constraints into the -lp format and then just call
> the subprocess module to run the program and then parse the output into a
> solution set of variables.  In fact, I've written this already to work out a
> variety of problems and even parsed in the dual solution, etc. for further
> analysis.
>
> lp_solve is about as efficient as cvxopt (based on highly-optimized C and
> Fortran routines) and is hard to beat among most of the AMPL-based solvers.
> I've tried a bunch of them (LOQO, SNOPT, KNITRO, etc.) and lp_solve is about
> as fast as you can get, for linear programs.
>
> -Michael Yang
>
>
> On Fri, Nov 1, 2013 at 8:16 AM, alex <argriffi at ncsu.edu> wrote:
>>
>> > If anyone has a source for problems involving at least dozens of
>> > variables,
>> > I'd love to try it out.
>>
>> This page has benchmarking information and test cases for linear
>> programming:
>> http://plato.asu.edu/ftp/lpcom.html
>> _______________________________________________
>> SciPy-Dev mailing list
>> SciPy-Dev at scipy.org
>> http://mail.scipy.org/mailman/listinfo/scipy-dev
>
>
>
> _______________________________________________
> SciPy-Dev mailing list
> SciPy-Dev at scipy.org
> http://mail.scipy.org/mailman/listinfo/scipy-dev
>