Disabling variable prescaling

回答済み

Jakub Komárek

2022年04月27日 18:13 編集済み

• 
• 

Hello,

as a part of some academic research I perform some (computationally intensive) rescaling of variables of a LP. I would now like to observe what effect (if any) this has on the time Gurobi needs to solve the problem.

However, in order for this to have any sense, Gurobi should not rescale the variables by itself, as that would render any previous rescalings futile. I have not managed to find any documentation relevant to if and when Gurobi performs such optimizations.

Is there a way to configure Gurobi to avoid any variable prescalings? In particular, is it enough to set the presolve level to 0?

I will be thankful for any insights!

0

• 

  David Torres Sanchez

  • Gurobi Staff

  2022年04月28日 12:47 編集済み

  Hi Jakub,
  
  If your variable scaling is better than those found in presolve then I would've thought that these will be kept.
  You can check this by looking a the presolved model. Using the Python API:

  m = Model()
  # Build and perform scaling
  p = m.presolve()
  p.write("presolved.lp")

  Check through that file and see if your variable scaling is altered. If this is the case you can go for disabling presolve altogether: \(\texttt{Presolve}=0\).
  
  Cheers, 
  David

  1

サインインしてコメントを残してください。