What's the meaning of "Learned Cut" in log

Answered

JIANSHU LI

• March 02, 2021 02:07

Dear Gurobi,

I found that the gurobi report applied "Learned Cut"（less than 3 for most of the time) on some of my models, but I can't find any reference about this kind of cut, and there is no parameter available for me to test its effect on my model. What's the background of this kind of cut? What type of problem does it apply to? How to tune it individually? Could you give me some reference about it？

Thank a lot!

0

• 

  Jaromił Najman

  Gurobi Staff

  • March 10, 2021 15:16
  • Edited

  Hi,

  During model presolve and manipulation of rows/columns, we call a "learned cut" a cut, which has been found incidentally while executing some presolve procedure. In this case the word "learned" rather stands for "remember", because we then remember these additional cuts and add them to the problem. These cuts can theoretically be found in any type of problem as they are not specific to any model structure. You cannot tune these cuts individually, i.e., you cannot tell Gurobi to find more of these cuts. There is no reference about this type of cuts.

  Best regards,
  Jaromił

  0
• 

  JIANSHU LI

  • March 19, 2021 15:25

  

  

  

  Got it, thank you very much  : ) 

  0

Please sign in to leave a comment.