Same optimization problem can be solved in MATLAB, but not in C#

Ongoing

Jianing Ding

• November 03, 2023 03:54
• Edited

I am transferring a mixed integer optimization program from MATLAB to C#, the MATLAB code is written as:

and the C# version:

With the same input parameters, this function can be successfully solved in MATLAB, but can't be dealt in C#. I've checked my code and there seems nothing wrong. What difference causes this?

Related to

• .NET
• MATLAB

0

• 

  Michel Soares

  • October 27, 2023 13:09

  

  

  Hi Jianing Ding,

  I would recommend you export a LP file in both codes and compare them, it might show you differences in the model (C#, MATLAB).

  I am not sure by what you mean with "can't be dealt in C#". Does the code not run at all? Is the model infeasible? What happens?

   

  0
• 

  Gwyneth Butera

  Gurobi Staff

  • October 27, 2023 14:06

  Can you check the bounds of the variables? I believe CVX does not use a lower bound of 0 while Gurobi does.

  0
• 

  Jianing Ding

  • October 28, 2023 07:12
  • Edited

  

  

  

  Hi Michel,

  Thanks a lot for you suggestion. But since I'm using both Gurobi and CVX in MATLAB, it seems that I can't export the LP file directly. Can you give me some advice?

  The C# function can run, but when using "model.Optimize()", it becomes infeasible, i.e., the model.Status equals to 4.

  0
• 

  Jianing Ding

  • October 28, 2023 04:18

  

  

  

  Hi Gwyneth,

  Thank you very much and you're right, Gurobi has a default lower bound of 0, and my constraints will result in negative values of variable u.
  I set the bounds of variable u to -GRB.INFINITY and GRB.INFINITY, the model.Status is still 4. Then I change this to -1000 and 1000, meanwhile I set the bounds of variable alpha(it's binary) to 0 and 1, model.Status becomes 3. Is there any other reason? 

  And I wonder why I need to set the bounds of binary variables, model.Status=4 without this additional constraint.

  0

Please sign in to leave a comment.