How to obtain the coefficient matrices for a QCQP problem?

回答済み

Canqi Yao

2025年07月08日 09:15 編集済み

• 
• 
• 

Hi,

I wanna obtain the specific coefficient matrices/vectors for a QCQP problem, e.g., QMatrix, QLMatrix, QObjective and QRHS in the attached pic. Is there any API for me to leverage for the coefficient matrices/vectors mentioned? 

 

Thanks

 

 

 

 

関連

• Attribute

0

• 

  Riley Clement

  • Gurobi Staff

  2025年10月12日 23:40 編集済み

  Hi Canqi,

  Edit: In gurobipy v13 the following functions are introduced: Model.getQ and Model.getQCMatrices.  For earlier versions, see the original response below.

  There isn't anything in our API to retrieve this data in matrix form, you would have to retrieve it in our “term based” API and construct the matrix, e.g. for your model m:

  import scipy.sparse as ss
  
  def get_q_matrix(model, qlinexpr):
      coeffs = [qlinexpr.getCoeff(i) for i in range(qlinexpr.size())]
      row = [qlinexpr.getVar1(i).index for i in range(qlinexpr.size())]
      col = [qlinexpr.getVar2(i).index for i in range(qlinexpr.size())]
      arr = ss.coo_array((coeffs, (row, col)), shape=(model.NumVars, model.NumVars))
      return (arr + arr.transpose())/2
  
  
  get_q_matrix(m, m.getObjective()) # QObjective
  get_q_matrix(m, m.getQCRow(m.getQConstrs()[0])) # QMatrix for the first quadratic constraint

  However perhaps you don't really need this data in matrix form, e.g. if your looking to find variables or constraints which lead to the extremities of the reported ranges then you can just work with the term-based API without needing to construct matrices.

  - Riley

  0

サインインしてコメントを残してください。