[Cbc] Cbc hanging

Sun Nov 1 22:07:42 EST 2009

Hi again John,

As I mentioned before this solution works well, thanks again. Total runtime is around 300 seconds.
However, when I went to implement this solution in my TP application I found that the runtime drastically increased to over 2 hours.

After some investigation I found that the solution found by RENS in the TP application was not within the 0.01% gap, but it is when I use driver4 with the settings below.
The only difference I could find was that in my TP application the problem is modelled as a maximization but because the driver1 program reads an mps file containing the problem, this is already a minimization.
When I negated the all the objective function coefficients and made the original problem a minimization in the TP application itself, the runtime was back down to about 300 seconds.

Regards,
Michael De Feyter

________________________________
From: John J Forrest [mailto:jjforre at us.ibm.com]
Sent: Thursday, 29 October 2009 11:04 PM
To: Defeyter, Michael
Subject: Re: [Cbc] Cbc hanging

Michael,

I don't think it is hanging - just wasting time doing cuts and strong branching.

One problem is that the heuristics don't seem to work well. There are two which have a chance - rounding and one called RENS. Neither worked with tp.mps. However I noticed that you have pairs of identical rows which are >=0 and <=0. There is code in Cbc to find identical rows, but I see that it only looks for identical rhs or where one is inside the other. If I manually make all such rows ==0, then rounding and RENS both work well. Cplex has a default to stop if the gap is less than 0.01%. If I do the same in Cbc it solves in two minutes.

I have fixed trunk to do something for intersecting identical rows. Very oddly, rounding still does not find anything - but RENS did.

If you look at driver4.cpp - you can see how to use command line version. With trunk I used

cbc tp.mps -divingc off -rens on -cuts off -ratio 0.0001 -trust 1 -strong 1 -hot 10 -solve

The -trust... stuff is just to stop it spending much time in strong branching.

John Forrest

[cid:304255102 at 02112009-0BDC]"Defeyter, Michael" ---10/28/2009 12:27:10 AM---Hi John, I've been testing Cbc with Clp to solve transport assignment problem, normally I use Cplex.

From:
"Defeyter, Michael" <Michael.Defeyter at bluescopesteel.com>

To:
"cbc at list.coin-or.org" <cbc at list.coin-or.org>

Date:
10/28/2009 12:27 AM

Subject:
[Cbc] Cbc hanging

Sent by:
cbc-bounces at list.coin-or.org

________________________________

Hi John,

I've been testing Cbc with Clp to solve transport assignment problem, normally I use Cplex.
Cbc works fine on smaller versions of the problem but when I use it on the full size problem, it seems to hang after a while.
The same problem solves fine using Cplex.
You can download the mps file from this location (it's a 2.6Mb zip file, I was not sure how to upload it to the mailing list so I've put it on my website):
http://www.flightdeckce.com/Download/tp.zip

I solve this using the sample2 code with the preprocess option turned on, when I don't use the preprocess option it solves for a bit longer but eventually hangs as well.
Thanks in advance,

Regards,
Michael De Feyter

________________________________

NOTICE - This message and any attached files may contain information that is confidential, legally privileged or proprietary. It is intended only for use by the intended recipient. If you are not the intended recipient or the person responsible for delivering the message to the intended recipient, be advised that you have received this message in error. Any dissemination, copying, use or re-transmission of this message or attachment, or the disclosure of any information therein, is strictly forbidden. BlueScope Steel Limited does not represent or guarantee that this message or attachment is free of errors, virus or interference.

If you have received this message in error please notify the sender immediately and delete the message. Any views expressed in this email are not necessarily the views of BlueScope Steel Limited._______________________________________________
Cbc mailing list
Cbc at list.coin-or.org
http://list.coin-or.org/mailman/listinfo/cbc

________________________________

NOTICE - This message and any attached files may contain information that is confidential, legally privileged or proprietary. It is intended only for use by the intended recipient. If you are not the intended recipient or the person responsible for delivering the message to the intended recipient, be advised that you have received this message in error. Any dissemination, copying, use or re-transmission of this message or attachment, or the disclosure of any information therein, is strictly forbidden. BlueScope Steel Limited does not represent or guarantee that this message or attachment is free of errors, virus or interference.

If you have received this message in error please notify the sender immediately and delete the message. Any views expressed in this email are not necessarily the views of BlueScope Steel Limited.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://list.coin-or.org/pipermail/cbc/attachments/20091102/2749f02a/attachment.html 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: graycol.gif
Type: image/gif
Size: 105 bytes
Desc: graycol.gif
Url : http://list.coin-or.org/pipermail/cbc/attachments/20091102/2749f02a/attachment.gif 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ecblank.gif
Type: image/gif
Size: 45 bytes
Desc: ecblank.gif
Url : http://list.coin-or.org/pipermail/cbc/attachments/20091102/2749f02a/attachment-0001.gif 