Friday, March 12, 2021

Some new/unknown LP/MIP solvers

Here are a few LP/MIP solvers that are (somewhat) new to me:

Google's GLOP

This is an LP solver and part of OR-Tools [1]. The license is Apache v2. 


An LP+MIP solver under active development by Julian Hall e.a.[2]

"HiGHS is a high performance serial and parallel solver for large-scale sparse linear programming (LP) and mixed-integer programming (MIP) models developed in C++11, with interfaces to C, C#, FORTRAN, Julia and Python."
MIT license, Successor to Clp?[3] Interesting to see the reference to John Forrest's "Idiot Crash".

This solver is also available through scipy.optimize.linprog [8].  Note that using that interface for any non-trivial model is somewhat insane.

MindOpt, Copt 

I don't know much about these two solvers[4,5]. They are commercial and both are from China. 
It is great to see some optimization tools from Asia. Until now things were largely US- and Europe-based.


There is also an LP/MIP solver that seems to be no longer available[6]. That is unfortunate as some users showed good performance with this solver. It was said to be open source, but that was a bit of a misnomer: just a few little peripheral pieces were available as source code. One can always hope that the availability and open-source licensing issues can be addressed in the future.


No comments:

Post a Comment