Adding a constraint with multiple MVar

Answered

Andreas Morr

• July 15, 2021 15:32

Hello all,

I am trying to create a constraint for my model with multiple MVar vector variables (1-D). Here is a minimal code example for which I hope you can reproduce the error I get:

vec_1 = model.addMVar(3)
vec_2 = model.addMVar(3)
const_vec = np.array([4,5,3])
model.addConstr(vec_1 <= np.multiply(const_vec, vec_2))

In essence, I want Gurobi to compare each entry of vec_1 with the entry of const_vec times the entry of vec_2. I realize that Gurobi doesn't like arithmetic inside of the constraint initialization, so I tried this:

a = np.multiply(const_vec, vec_2)
model.addConstr(vec_1 <= a)

Now a is of the type MLinExpr but I can not find a correct way to deal with this kind of object in the documentation. addConstr seems to be the wrong method. Maybe addLConstr?

Thanks in advance for the help!

Andreas

0

• 

  Eli Towle

  Gurobi Staff

  • July 16, 2021 04:05
  • Edited

  You could convert the constant vector to a (sparse) diagonal matrix, then build the constraint using Gurobi's support for matrix multiplication:

  import gurobipy as gp
  import numpy as np
  import scipy.sparse as sp
  
  m = gp.Model()
  
  x = m.addMVar(3, name='x')
  y = m.addMVar(3, name='y')
  
  a = np.array([4, 5, 3])
  m.addConstr(x <= sp.diags(a) @ y)

  0
• 

  Andreas Morr

  • July 27, 2021 08:22

  

  

  

  Thanks for the great answer! Works perfectly.

  0

Please sign in to leave a comment.