Decision Optimization

Hi all,

Suppose we have an MIP model coded in Python, and we solve it by CPLEX. When the solution procedure is over, I like to resolve the model using a different objective function. Is there a way or syntax to handle this? 

If yes, the second problem is being solved from scratch or just being reoptimized? 


Thanks

------------------------------
Amin
------------------------------

#DecisionOptimization

Hi

in

https://www.linkedin.com/pulse/making-optimization-simple-python-alex-fleischer

see

https://github.com/AlexFleischerParis/zoodocplex/blob/master/zooobjectivefunction.py

regards

------------------------------
ALEX FLEISCHER
------------------------------

Original Message:
Sent: Sat August 22, 2020 03:14 AM
From: Amin
Subject: change IP objective and resolve

Hi all,

Suppose we have an MIP model coded in Python, and we solve it by CPLEX. When the solution procedure is over, I like to resolve the model using a different objective function. Is there a way or syntax to handle this? 

If yes, the second problem is being solved from scratch or just being reoptimized? 


Thanks

------------------------------
Amin
------------------------------

#DecisionOptimization

You should specify whether you use docplex or the CPLEX Python API because the syntax is different between the two. Alex's answer is for docplex.

Yes, the second problem is solved from scratch. There is no way around this because changing the objective is likely to invalidate certain presolve reductions and for sure requires a new tree search. However, since changing the objective only affects optimality and not feasibility, CPLEX will use every solution it found in the first round as MIP start for the second round. This may or may not speed up the second solve.

------------------------------
Daniel Junglas
------------------------------

Original Message:
Sent: Sat August 22, 2020 09:34 AM
From: ALEX FLEISCHER
Subject: change IP objective and resolve

Hi

in

https://www.linkedin.com/pulse/making-optimization-simple-python-alex-fleischer

see

https://github.com/AlexFleischerParis/zoodocplex/blob/master/zooobjectivefunction.py

regards

------------------------------
ALEX FLEISCHER

Original Message:
Sent: Sat August 22, 2020 03:14 AM
From: Amin
Subject: change IP objective and resolve

Hi all,

Suppose we have an MIP model coded in Python, and we solve it by CPLEX. When the solution procedure is over, I like to resolve the model using a different objective function. Is there a way or syntax to handle this? 

If yes, the second problem is being solved from scratch or just being reoptimized? 


Thanks

------------------------------
Amin
------------------------------

#DecisionOptimization

Thanks Daniel!

I use CPLEX Python API.

------------------------------
Amin
------------------------------

Original Message:
Sent: Mon August 24, 2020 12:50 AM
From: Daniel Junglas
Subject: change IP objective and resolve

You should specify whether you use docplex or the CPLEX Python API because the syntax is different between the two. Alex's answer is for docplex.

Yes, the second problem is solved from scratch. There is no way around this because changing the objective is likely to invalidate certain presolve reductions and for sure requires a new tree search. However, since changing the objective only affects optimality and not feasibility, CPLEX will use every solution it found in the first round as MIP start for the second round. This may or may not speed up the second solve.

------------------------------
Daniel Junglas

Original Message:
Sent: Sat August 22, 2020 09:34 AM
From: ALEX FLEISCHER
Subject: change IP objective and resolve

Hi

in

https://www.linkedin.com/pulse/making-optimization-simple-python-alex-fleischer

see

https://github.com/AlexFleischerParis/zoodocplex/blob/master/zooobjectivefunction.py

regards

------------------------------
ALEX FLEISCHER

Original Message:
Sent: Sat August 22, 2020 03:14 AM
From: Amin
Subject: change IP objective and resolve

Hi all,

Suppose we have an MIP model coded in Python, and we solve it by CPLEX. When the solution procedure is over, I like to resolve the model using a different objective function. Is there a way or syntax to handle this? 

If yes, the second problem is being solved from scratch or just being reoptimized? 


Thanks

------------------------------
Amin
------------------------------

#DecisionOptimization

In the CPLEX Python API use Cplex.objective.set_linear() and Cplex.objective.set_sense() to modify the objective function.

------------------------------
Daniel Junglas
------------------------------

Original Message:
Sent: Tue August 25, 2020 05:40 PM
From: Amin
Subject: change IP objective and resolve

Thanks Daniel!

I use CPLEX Python API.

------------------------------
Amin

Original Message:
Sent: Mon August 24, 2020 12:50 AM
From: Daniel Junglas
Subject: change IP objective and resolve

You should specify whether you use docplex or the CPLEX Python API because the syntax is different between the two. Alex's answer is for docplex.

Yes, the second problem is solved from scratch. There is no way around this because changing the objective is likely to invalidate certain presolve reductions and for sure requires a new tree search. However, since changing the objective only affects optimality and not feasibility, CPLEX will use every solution it found in the first round as MIP start for the second round. This may or may not speed up the second solve.

------------------------------
Daniel Junglas

Original Message:
Sent: Sat August 22, 2020 09:34 AM
From: ALEX FLEISCHER
Subject: change IP objective and resolve

Hi

in

https://www.linkedin.com/pulse/making-optimization-simple-python-alex-fleischer

see

https://github.com/AlexFleischerParis/zoodocplex/blob/master/zooobjectivefunction.py

regards

------------------------------
ALEX FLEISCHER

Original Message:
Sent: Sat August 22, 2020 03:14 AM
From: Amin
Subject: change IP objective and resolve

Hi all,

Suppose we have an MIP model coded in Python, and we solve it by CPLEX. When the solution procedure is over, I like to resolve the model using a different objective function. Is there a way or syntax to handle this? 

If yes, the second problem is being solved from scratch or just being reoptimized? 


Thanks

------------------------------
Amin
------------------------------

#DecisionOptimization