CODEWITHSUNDEEP
python-2.7optimizationbinaryinteger-programming
John
John

Reputation: 33

Python - The integer linear programming (ILP) function in CVXOPT is not generating correct results

I'm trying to solve the simple example found in https://en.wikipedia.org/wiki/Integer_programming#Example using the CVXOPT library on Python 2.7 ; The optimal answers are either (1,2) or (2,2). I'm getting (0.0 , 0.0). What am I doing wrong in the code below ? Thanks !

import numpy as np
import cvxopt
from cvxopt import glpk

c=cvxopt.matrix([0,-1]) #-1 since we're maximising the 2nd variable
G=cvxopt.matrix([[-1,1],[3,2],[2,3],[-1,0],[0,-1]],tc='d')
h=cvxopt.matrix([1,12,12,0,0],tc='d')
(status, x)=glpk.ilp(c,G.T,h,B=set([0,1]))
print status
print x[0],x[1]    #should be (1,2) or (2,2)
print sum(c.T*x)

Upvotes: 3

Views: 6615

Answers (2)

John F
John F

Reputation: 317

Update for Python 3.8.8

from cvxopt.glpk import ilp
import numpy as np
from cvxopt import matrix
c=matrix([0,-1],tc='d')
G=matrix([[-1,1],[3,2],[2,3],[-1,0],[0,-1]],tc='d')
h=matrix([1,12,12,0,0],tc='d')
(status, x)=ilp(c,G.T,h,I=set([0,1]))
print (status)
print (x[0],x[1])
print (sum(c.T*x))

Calling:

cvxopt.glpk.ilp(c,G.T,h,I=set([0,1]))

returns: module 'cvxopt' has no attribute 'glpk'

Upvotes: 0

David
David

Reputation: 1999

Your code is basically correct but needs two minor modifications:

  1. The c-vector must be a double, too.
  2. The variables x[0] and x[1] are supposed to be integer, not binary.

Then, a working solution is given by:

import numpy as np
import cvxopt

c=cvxopt.matrix([0,-1],tc='d')
G=cvxopt.matrix([[-1,1],[3,2],[2,3],[-1,0],[0,-1]],tc='d')
h=cvxopt.matrix([1,12,12,0,0],tc='d')
(status, x)=cvxopt.glpk.ilp(c,G.T,h,I=set([0,1]))
print status
print x[0],x[1] 
print sum(c.T*x)

Upvotes: 6

Related Questions