Decision Optimization

Decision Optimization

Delivers prescriptive analytics capabilities and decision intelligence to improve decision-making.


#Analytics
#DecisionOptimization
#DecisionOptimization
 View Only
  • 1.  Cannot explain search termination

    Posted 10/24/12 02:06 PM

    Originally posted by: Vgoel77


    I am running CPO 12.4. As far as I know, the only limit I specify is a TimeLimit of 300 sec. For a couple of data sets, I observe that the search is terminated by limit and the log at the end of the solve says that it has actually spent 300 sec. But the last print to log of Time, avg fail depth etc says time = 82.71 sec (see second piece of snippet below). Why? what was CPO doing for the remaining 218 sec? Here are snippets from the log.

    Header
    
    ! ---------------------------------------------------------------------------- ! Minimization problem - 26783 variables, 37401 constraints, 1 phase ! Preprocessing : 6971 extractables eliminated ! DefaultInferenceLevel = Low ! LogVerbosity         = Verbose ! LogPeriod            = 1 ! SearchType           = Restart ! Workers              = 1 ! TimeLimit            = 300 ! Initial process time : 0.68s (0.40s extraction + 0.28s propagation) !  . Log search space  : 149060.9 (before), 143415.3 (after) !  . Memory usage      : 97.9 MB (before), 148.0 MB (after) !  . Variables fixed   : 706 ! Using sequential search. ! ----------------------------------------------------------------------------
    


    End of main log and log-summary
    
    15299.4912436    65878         48          on d[87][16][0] ! Time = 82.71s, Average fail depth = 750, Memory usage = 289.1 MB !          Best Branches  Non-fixed            Branch decision 15299.4912436    65879         21          on d[13][8][0] 15299.4912436    65880         14        F on d[28][7][0] 15299.4912436    65881         14        F on d[28][7][0] 15299.4912436    65882         45          on d[13][8][0] 15299.4912436    65883         35        F on d[55][1][0] 15299.4912436    65884         39          on d[55][1][0] 15299.4912436    65885         21          on d[29][0][0] 15299.4912436    65886         14        F on d[28][7][0] 15299.4912436    65887         14        F on d[28][7][0] 15299.4912436    65888         36          on d[29][0][0] 15299.4912436    65889         29        F on d[28][7][0] 15299.4912436    65890         36        F        - 15299.4912436    65891      26027        F        - ! ---------------------------------------------------------------------------- ! Search terminated by limit, 2 solutions found. ! Best objective         : 15299.5 ! Number of branches     : 65891 ! Number of fails        : 23497 ! Total memory usage     : 321.6 MB (283.9 MB CP Optimizer + 37.7 MB Concert) ! Time spent in solve    : 301.79s (301.39s engine + 0.40s extraction) ! Search speed (br. / s) : 218.6 ! ----------------------------------------------------------------------------
    

    I find this surprising because this is very different from what I oserve for all other cases. There I find that the times reported in the main body of the log go to 299.xx sec before it terminates. That makes more sense.

    Can someone please help me understand? Is it just a matter of chance that the last 12-13 nodes took that much time?
    Thanks
    #CPOptimizer
    #DecisionOptimization


  • 2.  Re: Cannot explain search termination

    Posted 10/24/12 03:16 PM

    Originally posted by: rdumeur


    Dear Vgoel77,

    This is indeed curious but not impossible if, as you rightly suggest, the lastest decisions take time (i.e. they triggers many propagations) for the solver log is not displayed during propagation.

    Cheers,

    Renaud
    #CPOptimizer
    #DecisionOptimization