FarkasDual and FarkasProof

回答済み

ahmad alanaqreh

• 2021年05月18日 17:54

I am trying to apply benders decomposition for a specific problem, I have to add a cut constraint in each iteration, if the subproblem feasible I am adding an optimality cut by using Pi values of the subproblems constraints, if it's InfeasibleI am adding feasibility cut by using the FarkasDual values of the subproblems constraints, but actually FarkasDual is not giving me right value since I have the code on AMPL and it's giving different values, I read that if the problem is Infeasible I have to use FarkasProof, is that true?

I have tried to use the FarkasProof as follow 

SP_model.getConstrByName ( 'one' ).FarkasProof

but I for the following error :

 in gurobipy.__getattr
AttributeError: 'gurobipy.Constr' object has no attribute 'FarkasProof'

although I have set the InfUnbdInfo to be one 

SP_model.setParam ( "InfUnbdInfo", 1 )

 

1

• 正式なコメント

  

  Simranjit Kaur

  Gurobi Staff

  • 2024年12月06日 00:31

  This post is more than three years old. Some information may not be up to date. For current information, please check the Gurobi Documentation or Knowledge Base. If you need more help, please create a new post in the community forum. Or why not try our AI Gurobot?.
• 

  Jaromił Najman

  Gurobi Staff

  • 2021年05月19日 08:47
  • 編集済み

  Hi Ahmad,

  FarkasProof is a model attribute. Thus, you have to access it via

  SP_model.FarkasProof

  Best regards,
  Jaromił

  1
• 

  ahmad alanaqreh

  • 2021年05月19日 08:34

  

  

  

  Hi Jaromil 

  Thanks for your help, so what  I don't understand how I can get the dual unbounded of a specific constraint for an infeasible model??

  in AMPL it is simply :

  constraint.dunbdd

  but it is not clear in python 

  0
• 

  Jaromił Najman

  Gurobi Staff

  • 2021年05月19日 08:50
  • 編集済み

  Hi Ahmad,

  I was not able to find a documentation on AMPL's \(\texttt{dunbdd}\) attribute.

  You can get the Farkas Dual for infeasibility certificate by

  SP_model.getConstrByName('one').FarkasDual

  Is this what you are looking for?

  Or are you looking for an unbounded ray? This can be obtained from the variable attribute UnbdRay

  SP_model.getVarByName('x').UnbdRay

  Best regards,
  Jaromił

   

  1
• 

  ahmad alanaqreh

  • 2021年05月19日 10:09

  

  

  

  Hi Jaromil 

  No, I am searching for Farkas Dual because the problem is infeasible, this unbdd is used to get the dual of the constraints if the problem infeasible(called "unbounded dual"):

  in AMPL if the problem is infeasible you can simply specify 

  suffix dunbdd

  then for the constraint 

  constraint.dunbdd

  but this dubdd(in AMPL) and FarkasDual(in python) gave totally different values although the constraint exactly the same !!

  0
• 

  Jaromił Najman

  Gurobi Staff

  • 2021年05月19日 11:15

  Hi Ahmad,

  Farkas certificate is not unique. Thus, it is very well possible that two different softwares return different certificates. As long as both are valid, there is no reason to worry.

  Best regards,
  Jaromił

  1
• 

  ahmad alanaqreh

  • 2021年05月19日 11:21

  

  

  

  Hi Jaromil 

  That is totally true, moreover, I found that FarkasDual is giving good values but the difference was the sign, I mean if the

  constraint.FarkasDual

  is positive 

  constraint.dunbdd

  is negative, so I just changed the sign and everything seems to be fine. 

  Thanks a lot for your help 

  0

投稿コメントは受け付けていません。