Interactions between partition heuristic and presolve

Answered

Thomas Leveringhaus

• February 03, 2023 09:56

Dear Gurubi Team,

if I have two variables x1 & x2 (and others) in my model and x1 and x2 have different attributes with respect to the partition heuristic. Additionally, if I have a constraint (or a set of constraints) that equals both variables or otherwise results in an elimination of one of the variables during presolve. Which partition attribute is retained in the model?

Thanks

Thomas

0

• 

  Jaromił Najman

  Gurobi Staff

  • February 13, 2023 11:27

  Hi Thomas,

  Please excuse getting to you only now.

  Which partition attribute is retained in the model?

  Gurobi keeps the partition attribute value of the variable surviving the presolve process. If the presolve process leads to x1 & x2 being both eliminated or replaced by a new variable all together, then the partition attribute is reset to its default value for the newly introduced presolve variable.

  Best regards, 
  Jaromił

   

  0
• 

  Thomas Leveringhaus

  • February 13, 2023 18:12

  

  

  

  Hi Jaromił Najman,

  thank you very much for your answer, which is still helpfull for me.

  As far as I know, there is no way to control which variables survive the presolve process, so I can't predifine the behaviour of the partition heuristic in such special cases we discussed here.

  I can think of arguments that speak for keeping the partition attribute of the surviving variables equal to -1, 0 or one of the sub-MIP-values, if equal variables have those different partition attributes.
  I suspect presolve is so sophisticated that it would be too complicated to define general decision rules for these special cases or to allow the user additional configuration options.

  But now I have clarity and must take care to foresee such cases as far as possible and to assign the partition attributes in advance in a meaningful way.

  Thanks

  Thomas

  0

Please sign in to leave a comment.