Open Source MILP software for Python with user-friendly API to define the optimization problem

Question

Following the accepted answer to

Assignment problem where assignments must be done sequentially

I would like to write a Python script which can solve the problem defined there. It's a Mixed Integer Linear Programming problem, thus I need an Open Source MILP solver, and a modeler. The constraints are probably very simple for you guys, but since I'm not familiar with MILP, it would be great if the modeler had an intuitive, easy to use API. Can you indicate me a Python modeler and solver which I can use to solve the problem? If you could also help me describing the MILP problem in the question above, using the modeler API, that would be great.

Take a look at the answer to this question: I've formulated my optimization model; now what?. You'll get to know commercial and open-source solvers, each with a Python API. And if you're looking for packages that call open-source solvers, check out this question: Interface for Cbc - COIN-OR — EhsanK, Jan 27 '20 at 14:03
In my opinion PuLP is outstanding : https://pypi.org/project/PuLP/ — Kuifje, Jan 27 '20 at 20:56
All the commercial solves have Python interfaces. I think mosek is a good one since I build it. — ErlingMOSEK, Jan 28 '20 at 06:32
The MIPCL solver has a nice interface on python and easy to use. I think easier than other solvers with good performance. — A.Omidi, Jan 28 '20 at 08:16
I didn't know that for MIP codes, there is a separation between modeling language and solver. Also, I'm not interested in commercial software. I edited the question to make this clear. — DeltaIV, Jan 28 '20 at 13:01
HiGHS has the best truly open-source MIP solver, and a limited Python interface. — SparseRunner, Sep 17 '21 at 12:04

score 7 · Answer 1 · answered Jan 27 '20 at 20:47

7

I'm not a Python user, but I've heard good things about Pyomo, which is an open-source modeling language. In addition to Pyomo, you would need a solver program (from among those they support), but you would not need to worry about any syntax issues related to the solver.

answered Jan 27 '20 at 20:47

prubin

39,078
3
37
104

I second this, Pyomo is great and automatically connects to any solver with an ASL interface (which is most of the COIN-OR stuff). – Nikos Kazazakis Jan 28 '20 at 14:02

score 7 · Accepted Answer · answered Jan 27 '20 at 21:19

7

I find Pulp is extremely easy to use, versatile and has good performance. Check it out here: https://pythonhosted.org/PuLP/solvers.html

answered Jan 27 '20 at 21:19

CMichael

1,333
6
20

score 4 · Answer 3 · answered Jan 30 '20 at 18:18

A more recent alternative is Python-MIP. It has built-in support of COIN-OR CBC. In addition, right now it supports one commercial solver, namely Gurobi. According to its website, it provides access to advanced solver features like cut generation, lazy constraints, MIP starts and solution pools. Since Python-MIP is mentioned on the COIN-OR projects webpage, it seems that Python-MIP is going to be open source in the long run.

score 2 · Answer 4 · edited Jan 29 '20 at 07:41

2

As suggested above, PULP is one of the open source solver. Apart from this, the best MILP solver available in open source is preferably offered in SCIP optimization suite. I think Google OR tools are built using SCIP only. Moreover, you can refer to COIN-OR where researchers from ORMS society have maintained their projects.

edited Jan 29 '20 at 07:41

ə̷̶̸͇̘̜́̍͗̂̄︣͟

5,412
5
22
55

answered Jan 29 '20 at 01:27

Divyam Aggarwal

668
3
10

5

SCIP is not open source. You cannot use it freely for commercial use for instance. See the license agreement. – ErlingMOSEK Jan 29 '20 at 08:04
Pulp is actually not a solver but a modeling framework - will update my answer accordingly. – CMichael Jan 29 '20 at 19:53
The HiGHS MIP solver is now as good as SCIP, and is truly open source. – SparseRunner Sep 17 '21 at 12:05

Open Source MILP software for Python with user-friendly API to define the optimization problem

4 Answers4