What libraries should I use for linear programming in python?

Question

A quick search on "python linear programming" turns up a lot of hits (e.g. this one). Looking through them, I see a fair number of complaints about outdated dependencies, poor documentation, etc.

Can anybody recommend a headache-free (e.g. fast, well-documented, easy-to-install, clean API) linear programming library for python?

here's a few more ... http://www.scipy.org/Topical_Software – dfb Apr 26 '11 at 16:30 — dfb, Apr 26 '11 at 16:30

score 10 · Accepted Answer · answered May 05 '12 at 23:14

10

I'd recommend looking at PULP and/or Pyomo.

answered May 05 '12 at 23:14

solvingPuzzles

8,541
16
69
112

3

Are these suitable for large instances, lets say at least hundereds of variables and thousands of constraints. – Andreas Mueller Nov 07 '12 at 16:15
1

Good question. I'm not sure about how PULP and Pyomo scale for large problems. However, I can say that most of my friends who do huge optimization problems (mostly for integrated circuit layouts) use IBM CPLEX as their optimization solver. Based on a quick google search, there is a CPLEX API for Python. However, CPLEX isn't free for non-academic use. – solvingPuzzles Nov 08 '12 at 01:43
...I just checked your stackexchange profile, Andreas, and it looks like you're a graduate student. You can get a free CPLEX license. :) – solvingPuzzles Nov 08 '12 at 01:44
@solvingPuzzles, would you know offhand if either of these has an API to NumPy arrays, along the lines `lpmin( x, A, c, lo, hi )` ? – denis Jul 25 '13 at 15:20
Good question. I'm not really sure...I haven't hacked around with optimization stuff in a few months, so I'm a bit out of touch at this point. – solvingPuzzles Jul 25 '13 at 20:01

score 5 · Answer 2 · answered Jan 23 '13 at 18:05

cvxopt is written by Lieven Vandenberghe and some of his collaborators. (This is the same Vandenberghe of the widely used convex optimization textbook by Boyd and Vandenberghe.) It's a general convex conic programming solver, and uses an interior point method. On the plus side it's well-documented, has many examples, and is easy to use. I believe it scales fairly well, though not as well as commercial products like Xpress, Gurobi, or cplex.

Looks like there's a pull request to scipy containing a (pure python) linear programming implementation, though. So a linear programming solver could be in scipy in the future.

For the record, here is that PR https://github.com/scipy/scipy/pull/218 — astrojuanlu, Jun 20 '13 at 15:31

score 2 · Answer 3 · answered Jun 04 '11 at 04:21

2

You might look at PuLP from the coin-or set of libraries.

http://www.coin-or.org/projects/

answered Jun 04 '11 at 04:21

Andrew Prock

6,900
6
40
60

score 1 · Answer 4 · answered Sep 02 '13 at 08:50

1

You can also take a look at or-tools, which includes a wrapper around widely used linear solvers such as GLPK.

answered Sep 02 '13 at 08:50

simleo

2,775
22
23

score 1 · Answer 5 · answered Jun 11 '15 at 18:57

1

As of 2015, scipy includes a method to solve linear programming models directly through scipy.optimize.linprog. It uses the Simplex algorithm.

answered Jun 11 '15 at 18:57

igordsm

464
2
8

score 1 · Answer 6 · answered Apr 26 '11 at 19:15

1

I don't know what you are specifically trying to do, but NumPy/SciPy are the usually first places to look for anything math related in Python.

answered Apr 26 '11 at 19:15

guidoism

7,820
8
41
59

Numpy is an array library, with some extra functionality tossed in for backwards compatibility. Scipy has some optimization routines, but as of now I think it's only general non-linear solvers. Scipy does not currently have a solver specialized for linear programs. – cjordan1 Jan 23 '13 at 17:43

What libraries should I use for linear programming in python?

6 Answers6

Linked