BookmarkSubscribeRSS Feed
deleted_user
Not applicable
Hi all,

I have solved an mixed integer linear problem with PROC LP and PROC OPTMILP. The solution, however, are not equal. The PROC OPTMILP has found a better solution (with respect to the objective). Why? Can anybody tell me something about it? I first had the problem formulated in sparse, solved this with PROC LP. Then I converted the sparse in MPS format and solved it with PROC OPTMILP.

Thanks Message was edited by: Chavalit
9 REPLIES 9
YanXu
SAS Employee
PROC LP is not under active development anymore. PROC OPTMILP is faster and more robust.

If you send us the data, we will investigate this issue. Thanks!
deleted_user
Not applicable
Dear Yan Xu,

Thank you for the hint. I have been already in despairing, because of the PROC LP has issued different results, sometime worse than and sometime optimal as OPTMILP, sometimes even infeasible. I will try to ask for permission to send you the data. It will be on next week, but I did not see any way to put file on this forum. How can I send it to you?

I would now like to ask for more information. We work mainly with mixed integer linear problem, but sometimes also pure linear problem (without integer). The call to PROC ... is wrote in a SAS macro integrated into our application. We generate problem in sparse format and convert to mps format with MPSOUT of PROC LP. So I have the following questions:

1) Is there a way that SAS recognizes automatically if there is a mixed integer linear or pure linear problem and then give it to the appropriate procedure (OPTLP or OPTMILP)?

2) What happens when I solve a linear problem with OPTMILP or solve a mixed integer linear problem with OPTLP, would it yield the same results? Could you explain to me what are the pros and cons for it?

Thanks very much
YanXu
SAS Employee
1) I suggest you consider using OPTMODEL. It is a modelling language like AMPL. After creating model, OPTMODEL can choose solver automatically based on model type (LP, MILP, NLP, etc.)

http://support.sas.com/documentation/cdl/en/ormpug/59679/HTML/default/optmodel.htm


2) You can solve a pure LP by OPTMILP, the primal solution will be the same, but OPTMILP will not give dual solution and you don't have the option to use other LP solvers (Primal Simplex and Interior Point).

As the data, you can either put a place that I can download, or enter a request via SAS surport

http://support.sas.com/ctx/supportform/createForm

Note the product is SAS/OR. In the support request, say the data is for Yan Xu.

thanks!
deleted_user
Not applicable
Yan Xu,

It is quite expensive for us to change to OPTMODEL, but the information will be useful for later. Thanks for the tip! I think OPTMILP is currently the best choice for us.
Please look out for my sending of the data in a few days.

Thank you. I really appreciate your help.
deleted_user
Not applicable
Hi Yan Xu,

Sorry that I have still can not send the data. We believe rather that our application is not stable, so that not exactly same sparsedata generated. However, we still investigate. I will notify you as soon as we know more.
deleted_user
Not applicable
Dear Yan

I sent the data to you via technical support. After the investigation we are more or less sure that the same sparse are generated. Maybe you can help us.

Thanks
YanXu
SAS Employee
thanks for the data. Will get back to you soon.
tonya_sas
SAS Employee
Hello,

I am the technical support consultant for SAS/OR and have not received your reported problem or data through our tracking system. Will you please let me know your tracking number so that I can investigate?

If you do not have a tracking number, please email your data to support@sas.com. If you zip your data, you will need to preface the zip file with ZQJ_, for example ZQJ_test.zip to pass our firewall.

Thank you,
Tonya
deleted_user
Not applicable
Dear Tonya,

The tracking number is 7610393244. Thanks.

sas-innovate-2024.png

Join us for SAS Innovate April 16-19 at the Aria in Las Vegas. Bring the team and save big with our group pricing for a limited time only.

Pre-conference courses and tutorials are filling up fast and are always a sellout. Register today to reserve your seat.

 

Register now!

Multiple Linear Regression in SAS

Learn how to run multiple linear regression models with and without interactions, presented by SAS user Alex Chaplin.

Find more tutorials on the SAS Users YouTube channel.

Discussion stats
  • 9 replies
  • 1486 views
  • 0 likes
  • 3 in conversation