Complete different solution for nearly identical paramters
AnsweredHello, I have the following question. I have a model ("m") for planning shifts and in it a parameter that I can vary. Now, even if I change this now marginally, then completely different plans come out, although if one solves it manually, the same plan before for the marginally smaller value of the parameter must come out.
1) What is the reason for this?
2) Can I force it to change only when it is absolutely necessary?
-
-
Gurobi Staff
Hi Lorenz,
Changing any aspect of the model may lead to the solver using different "solution paths". Is this purely a feasibility problem or is there a non-zero objective function?
As an experiment I would try running your model with different values of Seed, and seeing if there are multiple different optimal solutions being produced.
What exactly is this model parameter you are changing? The answer will indicate whether the model can be warm started with previous solutions.
Can you explain exactly what the following means? I wasn't sure what is meant by solving it manually.
although if one solves it manually, the same plan before for the marginally smaller value of the parameter must come out
- Riley
0 -
-
Riley Clement Okay thank you. I tried different seed but it yields the same results.
Nevermind the last part. One another note, is there any quotable resource where I can read up the fact regarding the different solution paths?
0 -
-
-
Hi Lorenz,
I'd suggest the following could be a good resource:
Andrea Lodi, Andrea Tramontani (2014)
Performance Variability in Mixed-Integer Programming.
INFORMS TutORials in Operations Researchin particular the section "The roots of performance variability".
- Riley
0 -
Please sign in to leave a comment.
Comments
3 comments