MILP solution taking too long, while the incumbent obj value is at the optimal level from early on

Makedon Karasavvidis

• May 07, 2020 13:13

Using GUROBI with python.

When solving a MILP, I notice that the incumbent is very early (25sec) at the optimal point but the best bound is so slow to fall (maximization problem) that it takes ages (2000sec+) for reaching the optimal solution. This holds for varying input datasets so I can deduct that it is a general trend.

Any suggestions for changing the parameters that could make it happen faster? I have tried to change lots of them (cuts, MILPfocus etc.) and also use the model.tune() function.

0

• 

  Ismael Osuna

  • May 07, 2020 20:27

  

  

  Are your variables are expression bounded? maybe the solver is taking too much time discarding unrealistic values, e.g. for PWL functions, before applying sos constraints

  0
• 

  Makedon Karasavvidis

  • May 08, 2020 14:38

  

  

  

  Thank you very much for your response.

  I guess you mean continuous variables that are used (auxiliary) in big-M expressions ? By saying expression bounded do you mean something different than just bounded?

  Well I did that and saw a certain amount of improvement and thank you for that. But still the trend remains there, again the incumbent is at the optimal level a lot earlier than the finalization of the run. Any more suggestions please?

  0

Please sign in to leave a comment.