MIP problem - Bounds improves slowly

Answered

Farnaz Fallahi

• February 15, 2021 21:01

Hi all, 

I am running a MIP problem. Attached is the MIP log. As you see, bounds move very slowly. I changed the MIPFocus=3, but it didn't help. I would appreciate it if you can give me some suggestions. 

Thanks a lot

0

• 

  Jaromił Najman

  Gurobi Staff

  • February 16, 2021 10:18

  Hi Farnaz,

  Could you also share the model statistics, i.e., the first LOG lines stating the number of rows, columns, nonzeros and the coefficient ranges?

  To me, it looks like while the bound is moving slowly, it still does make some progress and it should be more worth it to try to focus on improving the feasible solution. Thus, MIPFocus=1 or 2 might be helpful. Trying the new NoRelHeurTime heuristic might help here as well. Note that this heuristic is only available in the latest 9.1.x versions.

  Best regards,
  Jaromił

  0
• 

  Farnaz Fallahi

  • February 16, 2021 18:39

  

  

  

  Hi Jaromil,

  Thanks for your quick response. I attached the model statistics. I tried MIPFOCUS= 1 and 2 per your suggestion. In both cases, it makes the convergence accelerated at the beginning but got stuck when the gap reaches 6% ish. 

  0
• 

  Jaromił Najman

  Gurobi Staff

  • February 17, 2021 08:52

  Hi Farnaz,

  There are a couple of things you should certainly try:

  • Try improving the ranges of your matrix coefficients. These currently have a range of 11 orders of magnitude, which may lead to severe numerical issues. Please have a look at our Guide to Numerical Issues for more insight.
  • Upgrade to the latest version 9.1.1
  • After upgrading to the latest version, try the no relaxation heuristic

  Best regards,
  Jaromił

  0

Please sign in to leave a comment.