How to use the function "model.relax()"?

Answered

Ariel W

• March 11, 2024 09:30

Using relax function of Gurobi can we derive the bound of original model? can it help me simplify the MIP model solving? how can i do this, please give me an example. Thank you!

Related to

• Root Node

0

• 

  Michel Soares

  • March 11, 2024 13:49

  

  

  Hi Ariel,

  The first thing Gurobi will do is solving the relaxed model (meaning that all integer/binary constraints are ignored). Afterwards, it will use heuristics and branch and bound to find integer feasible solutions. It might be worth it looking into how a solver optimizes a MIP for you to understand better what is the relaxed model and its use.

  I do not see how you could use model.relax to simplify your original model. The bounds of your model should also be easily available when solving the MIP directly.

  0
• 

  Ariel W

  • March 12, 2024 01:35

  

  

  

  Thanks Soares!

  I got what the meaning of this, but when i use the "model.relax()" in my program, it seems to ignore almost all constraints as my constraints are seemly all related to binary variables:)

  Are there any useful cases that I can learn and refer to?

  0
• 

  Michel Soares

  • March 12, 2024 12:44

  

  

  Model.relax is useful in very specific cases, it is not common to be used, I can think of one case:

  • There are some MIP models which might take too long to get to answer, requiring sophisticated user-defined-heuristics. In order to do this, you might want to optimize the relaxed model first and then use its value to either: set initial solution for the MIP, add constraints/cuts to model, set variable hints.

   

  0
• 

  Ariel W

  • March 13, 2024 01:32

  

  

  

  Thanks for your careful reply,Soares. I will find the way for optimizing my model based on your advice.

  0

Please sign in to leave a comment.