[Clp] Issue with Clp executable for maximization and for binary variables

[John Forrest]

Sourav,

Cbc reads arguments in order so you need to put -threads 6 before -solve.

Problem is big, but switching off preprocessing is probably counter 
productive.  Given those statistics, I would expect preprocessing to 
reduce size considerably.

If you want to send me the problem (compressed), I amy have time to take 
a quick look.

John Forrest
On 22/05/14 13:30, Sourav Garg wrote:
> Thanks John.
>
> Cbc was what I actually needed.
> I am using the standalone executable of cbc which solves the Binary 
> Integer program that I provide using a .mps file.
> For some instances I get the solution within seconds, but for one of 
> my inputs, it took one day to complete and get an optimal solution. 
> The corresponding stdout logs were like this:
>
> /"/
> /.../
> /Problem BIP has 287088 rows, 7346 columns and 509272 elements
> /
> /.../
> /.../
> /.../
> /
> /
> /Result - Optimal solution found/
> /
> /
> /Objective value:                0.00004017/
> /Enumerated nodes:               92760/
> /Total iterations:               85136638/
> /Time (CPU seconds):             85473.77/
> /Time (Wallclock seconds):       85683.41/
> /
> /
> /Total time (CPU seconds):       85474.08   (Wallclock seconds):       
> 85683.75/
> /
> /
> /"/
> Reason is probably the size of data.
> I realized that there is an option of using multiple threads while 
> running cbc (source: man page of Ubuntu), I tried to use that:
>
> $ cbc myFIle.mps solve -threads 6 -solution cbcSoln.txt
>
> /"
> /
> /.../
> /threads was changed from 0 to 6
> /
> /../
> "
>
> It gives confirmation of changing threads number, but I don't see any 
> of my other cores being used, just a single processor at 100%. (I am 
> using Dell Precision T5500 Processor: Intel® Xeon(R) CPU X5675 @ 
> 3.07GHz × 12, Memory: 47.2 GB).
> I want to reduce the time it takes for finding the solution. Along 
> with switching off the cuts and preprocess options to save time, can I 
> further change some more parameters so that I can save time and get an 
> optimal solution?
>
> Thanks,
> Sourav
>
>
> On Tue, May 20, 2014 at 7:08 PM, John Forrest 
> <john.forrest at fastercoin.com <mailto:john.forrest at fastercoin.com>> wrote:
>
>     Sourav,
>
>     1.  -maximize is just a flag - you should follow it with
>     -dualsimplex or similar.
>     2. Clp does not solve integer problems - Cbc does so download that.
>
>     John Forrest
>
>     On 20/05/14 10:42, Sourav Garg wrote:
>>     Hello,
>>
>>     I am using Clp1.15.6 on an Ubuntu 12.04 machine.
>>     I have compiled and installed the executable as mentioned in the
>>     readme install notes, which went successful.
>>
>>     1. I have been trying some linear programs using the Clp
>>     executable with a mps file. It gives me right solution for a
>>     minimization problem (which is the default direction of
>>     optimisation), but it doesn't give me correct solutions for
>>     maximisation problems.
>>
>>     For maximisation, I use it as follows:
>>     ./clp ./my.mps -maximize -solution solFile
>>
>>     What I get in the file is:
>>     status unknown
>>     objective value 0
>>     columns val 0 as well..
>>
>>     2. Apart from this, I have a Binary Integer Program problem, for
>>     minimisation, in which I pass all the bounds as BV in the mps file.
>>     But in the solution I get floating point values for the non zero
>>     columns instead of getting the value of 1. I wanted to know if
>>     Binary Integer Programming is supported with the library?
>>
>>     Thanks,
>>     Sourav
>>
>>
>>
>>     _______________________________________________
>>     Clp mailing list
>>     Clp at list.coin-or.org  <mailto:Clp at list.coin-or.org>
>>     http://list.coin-or.org/mailman/listinfo/clp
>
>
>     _______________________________________________
>     Clp mailing list
>     Clp at list.coin-or.org <mailto:Clp at list.coin-or.org>
>     http://list.coin-or.org/mailman/listinfo/clp
>
>
>
>
> _______________________________________________
> Clp mailing list
> Clp at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/clp

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/clp/attachments/20140522/279b6193/attachment.html>