Add additional constraints but the objective decreased in minimize problem
回答済み
Recently, I encountered an issue with Gurobi, which I suspect is related to precision. Here are the details:
If I don’t add these additional constraints, Gurobi returns an optimal solution of 1.187605836561e+06. If I add these constraints, Gurobi returns an optimal solution of 1.159162044630e+06.
The difference between these two solutions is about 2.5%, which is relatively significant for me.
We all know that for a minimization problem, adding constraints typically worsens the optimal solution. Although both of these solutions are deemed optimal, why does adding constraints lead to finding a better value?
Besides, I try to adjust FeasibilityTol and OptimalityTol to 10e-9, and solutions are getting better with 1.174222385819e+06. How to fix this problem and how to explain this issue to my reviewers? Many thanks!!!
-
Are you using the latest version of Gurobi (v11.0.3)?
0 -
Hello Jaromił Najman！Nice to meet you again and sorry for reply you so late because of busy days. Yes, I use the latest version 11.0.3.
0 -
Hi Runfeng, I hope you're doing well.
Could you please share the two models such that we can have a closer look? Note that uploading files in the Community Forum is not possible but we discuss an alternative in Posting to the Community Forum.
0 -
Hi Jaromił,
Here are the two models in .lp format.
https://1drv.ms/f/c/5fddeb3c4240ada2/EgjCx6zP4ShJn4K8aHhuiqcBunzcGxQLliQDHvG5ejcGUA?e=rpVtU3
These files are a bit large. I hope they will be helpful.
Besides, the additional constraints are part of the heuristic solution obtained by another program.
I want to compare these two methods and find this problem.If you need Python source code in the figure, I will send it to your email.
0 -
Hi Runfeng,
I was able to reproduce the issue on our side. However, I did not find the reason for the behavior yet. I will let you know once I have new information. It may however take a few days.
Best regards,
Jaromił0 -
Hi Jaromił,
Deeply thank you for your help. Looking forward to your good news.
0 -
Hi Runfeng,
Just wanted to let you know that we are still on it.
Best regards,
Jaromił0 -
Hi Jaromił,
It’s great to hear from you. I know that you have all been working. Identifying problems is often simple or happens by chance, but solving them is indeed a complex task. I look forward to hearing your good news.
Best regards,
Runfeng Yu0 -
Hi Runfeng,
Thank you for your patience. We were able to identify the source of the issue and it will be fixed with the upcoming version 12 release, which is planned for the end of the year.
Unfortunately, due to difficult numerics, there is no good workaround for the issue in version 11. You could try setting Cuts=0 but this may harm performance a lot.
Thank you for reporting this issue.
Best regards,
Jaromił0 -
Hi Jaromił,
I’m very pleased to hear your good news. As a Gurobi user, it’s an honor to have identified such an issue, and I’m delighted that it will be resolved in the next version. Gurobi is the best solver in the world because of the convenient opportunities to communicate with the staff and receive responses. Overall, I’m thrilled that the issue I discovered can be fixed, making Gurobi even better.
Best regards,
Runfeng Yu0
サインインしてコメントを残してください。
コメント
10件のコメント