[MATLAB] Modeling QCQP(Quadratically Constrained Quadratic Problems)

Answered

Yooseung Choi

• March 20, 2024 17:10

I'm trying to solve QCQP and trying to generate a model using MATLAB.

From 'MATLAB Reference Manual,' it says that for quadratic constraints, one can use 'quadcon' function which can describe x'*Qc*x + q'*x <= beta.

ex)

model.quadcon(i).Qc = sparse(Qj(-,-));   % Quadratic terms

model.quadcon(i).q = qj(-)';                     % Linear terms

model.quadcon(i).rhs= rhs(-);                    % Right hand side

model.quadcon(i).lhs= lhs(-);                     % Left hand side

Here's my question:

Gurobi uses 'model.A' for defining linear constraints. 

model.A = sparse([-]);

Would Gurobi recognize this as the same linear constraint if I define the linear constraint as

model.quadcon(i).q = some vector [-]

 

with model.quadcon(i).Qc as a matrix of zeros and enter model.A = sparse(zeros(-,-))?

 

0

• 

  Riley Clement

  Gurobi Staff

  • March 20, 2024 23:29

  Hi Yooseung,

  I don't have the working knowledge to answer this, maybe someone else will, but it seems like something you could easily test?

  You can write out an LP file with gurobi_write() and inspect it.

  - Riley

  1
• 

  Jaromił Najman

  Gurobi Staff

  • March 22, 2024 11:45

  Hi Yooseung,

  If I understand your post correctly, your concern is about defining a linear constraint via the quadcon struct. It should be no problem to define a linear constraint via the quadcon struct, i.e., set the \(Q\) matrix = 0. If you see some suspicious or wrong behavior, please let us know.

  As Riley suggested, you can always write out an LP file and analyze what is happening.

  Best regards, 
  Jaromił

  1

Please sign in to leave a comment.