Accessing Objective Q matrix in Gurobi Python Shell

Question

I'm attempting to reformulate an Objective Q matrix in an optimization problem (with quadratic constraints and a quadratic objective function) that I am solving with Gurobi and Python. Gurobi has the option of adding in constraints and objective functions as linear expressions instead of fiddling with matrices so I don't have the original matrix, Gurobi creates it for me via my objective equations and coefficients.

To perform analysis on convexity psd properties of the Objective Q matrix, I need to have the Q (objective matrix) and A (constraint matrix). Does anyone know if there is aa command in the gurobi.py shell that allows me to access/view the Objective Q matrix?

Thank you in advance and feel free to ask for clarification if necessary!

Answer 1

Gurobi Optimizer does not have a simple function to retrieve matrices from a Model object. However, you can get that data iteratively via LinExpr and QuadExpr objects. Here is a simple program that iterates through a linear constraint matrix (A) and prints the coefficients:

from gurobipy import *

m = read('afiro.mps')

for ct in m.getConstrs():
    row = m.getRow(ct)
    for i in range(row.size()):
        print("%s %s %f" % (ct.ConstrName, row.getVar(i).VarName, row.getCoeff(i)))

Adapt this to a QuadExpr for a quadratic objective or quadratic constraint.

Answer 2

With gurobipy 9.x, you can use the (undocumented) Model.getQ() method to get the objective Q matrix as scipy.sparse.coo_matrix :

from gurobipy import read

m = read('mwe_example.mps')
Q = m.getQ()
Qdense = Q.todense()