problem in reducing gap

369 views
Skip to first unread message

Luke Li

unread,
May 8, 2012, 10:37:02 PM5/8/12
to am...@googlegroups.com
Hi everyone, 

I have difficulty in reaching the optimum solution for my MIP problem. 
The record from cplex shows that the gap is at 65% and then cplex concludes the optimum solution is found.
I didn't put any limit on solving time ot the 'mipgap', but it just stops with such large gap.

Any suggestion to reduce the gap?

Thanks in advance.

Luke

################### Record from cplex  ########################
CPLEX 12.2.0.0: mipdisplay 2

mipinterval 100

MIP Presolve eliminated 7491 rows and 1070312 columns.

MIP Presolve modified 2931568 coefficients.

Reduced MIP has 3370 rows, 1438448 columns, and 7468221 nonzeros.

Reduced MIP has 12 binaries, 0 generals, 0 SOSs, and 0 indicators.

Probing fixed 3 vars, tightened 0 bounds.

Probing time =    0.89 sec.

Clique table members: 21.

MIP emphasis: balance optimality and feasibility.

MIP search method: dynamic search.

Parallel mode: deterministic, using up to 4 threads.

Root relaxation solution time =  262.53 sec.



        Nodes                                         Cuts/

   Node  Left     Objective  IInf  Best Integer     Best Node    ItCnt     Gap



      0     0  2.86770e+011     9                2.86770e+011       43         

      0     0  2.86770e+011     9                   Cuts: 942    83372         

      0     0  2.86770e+011     9                   Cuts: 942   170842         

Heuristic still looking.

Heuristic still looking.

Heuristic still looking.

Heuristic still looking.

      0     2  2.86770e+011     9                2.86770e+011   170842         

Elapsed real time = 46376.41 sec. (tree size =  0.01 MB)

*     9     5      integral     0  2.17674e+012  2.86770e+011   644968   86.83%

*    11     3      integral     0  8.35943e+011  2.86770e+011   766493   65.69%



Implied bound cuts applied:  927



Root node processing (before b&c):

  Real time             = 46354.31

Parallel b&c, 4 threads:

  Real time             = 13187.56

  Sync time (average)   = 3125.28

  Wait time (average)   = 8301.23

                          -------

Total (root+branch&cut) = 59541.87 sec.
#################################################################

Robert Fourer

unread,
May 10, 2012, 2:53:14 PM5/10/12
to am...@googlegroups.com

The optimal solution message that CPLEX generates at the end of its run (not shown in your posting) shows the number of nodes created.  Since your problem has only 12 binary variables, it is possible that fewer than 100 nodes were needed.  But since you have set the CPLEX option "mipinterval 100", only every 100th node is displayed in the CPLEX output (except for nodes 9 and 11 where an improved integer solution was been found).  Thus there might be a lot of nodes after node 11 where CPLEX is reducing the mipgap considerably; to get a better idea of what is happening, try setting "mipinterval 1" instead.

 

Bob Fourer

4...@ampl.com

 

 

From: am...@googlegroups.com [mailto:am...@googlegroups.com]

On Behalf Of Luke Li
Sent: Tuesday, May 08, 2012 9:37 PM
To: am...@googlegroups.com
Subject: [AMPL 5767] problem in reducing gap

--
You received this message because you are subscribed to the Google Groups "AMPL Modeling Language" group.
To view this discussion on the web visit https://groups.google.com/d/msg/ampl/-/srzolB2Mc5sJ.
To post to this group, send email to am...@googlegroups.com.
To unsubscribe from this group, send email to ampl+uns...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/ampl?hl=en.

Luke Li

unread,
May 14, 2012, 2:58:06 AM5/14/12
to am...@googlegroups.com, 4...@ampl.com
Thanks Bob.
I will try it out.

Regards
Luke

To unsubscribe from this group, send email to ampl+unsubscribe@googlegroups.com.

Reply all
Reply to author
Forward
0 new messages