Decision Optimization

Decision Optimization

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


#Analytics
#DecisionOptimization
#DecisionOptimization
 View Only
Expand all | Collapse all

Problem infeasible using java, but feasible using command line

  • 1.  Problem infeasible using java, but feasible using command line

    Posted 08/19/15 10:37 AM

    Originally posted by: FlorianKirchner


    Hi all,

    I have a problem solving a QP-Problem using the JAVA-API. To diagnose the problem, I exported a .mps file and used command line cplex. Surprisingly that worked fine.
    I attached the mps file to this post.

    What I could see from the output is, that the presolving has different results:

    Java:

    Number of nonzeros in lower triangle of Q = 240
    Using Approximate Minimum Degree ordering
    Total time for automatic ordering = 0.00 sec. (0.01 ticks)
    Summary statistics for factor of Q:
    Rows in Factor= 120
    Integer space required= 120
    Total non-zeros in factor = 360
    Total FP ops to factor= 1320
    
    Warning:diagonal perturbation of 5.6e-009 required to create PSD Q.
    
    Tried aggregator 1 time.
    QP Presolve eliminated 406 rows and 360 columns.
    Aggregator did 452 substitutions.
    Reduced QP has 308 rows, 509 columns, and 2436 nonzeros.
    Reduced QP objective Q matrix has 100 nonzeros.
    Presolve time = 0.00 sec. (1.15 ticks)
    Parallel mode: using up to 4 threads for barrier.
    
    ***NOTE: Found 3 dense columns.
    
    Number of nonzeros in lower triangle of A*A' = 4433
    Using Approximate Minimum Degree ordering
    Total time for automatic ordering = 0.00 sec. (0.31 ticks)
    Summary statistics for Cholesky factor:
    Threads = 4
    Rows in Factor= 311
    Integer space required= 1052
    Total non-zeros in factor = 8251
    Total FP ops to factor= 312689
     ItnPrimal ObjDual ObjPrim Inf Upper InfDual Inf
    
     0 -4.4394278e+010 -1.4596028e+005 4.66e+002 2.00e+000 2.85e+012
     1 -5.9186635e+011 -3.2525689e+008 4.12e+002 1.70e+000 2.42e+012
     2 -2.3775642e+010 -1.3257318e+009 3.50e+002 1.39e+000 1.98e+012
    [...]
    55 -1.0079217e+009 -1.0079899e+009 4.53e-009 3.39e-021 3.94e-001
    *-1.0079271e+009 -1.0079587e+009 7.77e-009 8.47e-021 9.49e-001
    Barrier time = 0.05 sec. (12.19 ticks)
    
    Total time on 4 threads = 0.05 sec. (12.19 ticks)
    

     


    Command Line:
     



    Number of nonzeros in lower triangle of Q = 240
    Using Approximate Minimum Degree ordering
    Total time for automatic ordering = 0.02 sec. (0.01 ticks)
    Summary statistics for factor of Q:
      Rows in Factor            = 120
      Integer space required    = 120
      Total non-zeros in factor = 360
      Total FP ops to factor    = 1320
    
    Warning:  diagonal perturbation of 3.0e-007 required to create PSD Q.
    
    Tried aggregator 1 time.
    QP Presolve eliminated 398 rows and 352 columns.
    Aggregator did 452 substitutions.
    Reduced QP has 316 rows, 517 columns, and 2452 nonzeros.
    Reduced QP objective Q matrix has 108 nonzeros.
    Presolve time = 0.02 sec. (1.15 ticks)
    Parallel mode: using up to 4 threads for barrier.
    
    ***NOTE: Found 3 dense columns.
    
    Number of nonzeros in lower triangle of A*A' = 4475
    Using Approximate Minimum Degree ordering
    Total time for automatic ordering = 0.00 sec. (0.31 ticks)
    Summary statistics for Cholesky factor:
      Threads                   = 4
      Rows in Factor            = 319
      Integer space required    = 1060
      Total non-zeros in factor = 8325
      Total FP ops to factor    = 313417
     Itn      Primal Obj        Dual Obj  Prim Inf Upper Inf  Dual Inf
       0 -4.4394278e+010 -1.4596028e+005 4.66e+002 2.00e+000 2.85e+012
       1  4.4606660e+011 -1.5599676e+009 3.16e+002 1.27e+000 1.80e+012
       2  2.4921306e+012 -2.4223290e+009 3.56e+002 1.15e+000 1.64e+012
       3  1.2774827e+013 -1.4150052e+011 4.95e+002 1.04e+000 1.48e+012
       4  2.3005673e+013 -4.1096898e+010 8.30e+002 6.99e-001 9.96e+011
       5  2.4135004e+013 -3.4135482e+010 9.31e+002 5.56e-001 7.92e+011
      [...]  
      33 -1.0079333e+009 -1.0079334e+009 5.47e-010 0.00e+000 3.19e-001
      34 -1.0079333e+009 -1.0079333e+009 1.16e-010 1.69e-021 4.47e-001
      35 -1.0079333e+009 -1.0079333e+009 1.06e-010 0.00e+000 1.84e-001
    Barrier time = 0.09 sec. (8.99 ticks)
    
    Total time on 4 threads = 0.09 sec. (8.99 ticks)
    
    Barrier - Optimal:  Objective = -1.0079333106e+009
    Solution time =    0.09 sec.  Iterations = 35
    Deterministic time = 8.99 ticks  (95.63 ticks/sec)
    


    Does anyone have an idea why this is happening?

    Thanks,
    Florian


    #CPLEXOptimizers
    #DecisionOptimization


  • 2.  Re: Problem infeasible using java, but feasible using command line

    Posted 08/19/15 11:21 AM

    Originally posted by: FlorianKirchner


    I found the solution myself: 

    Exporting to sav format is closer to the original model than mps. When I export to sav, the problem is infeasible on the command line all well.


    #CPLEXOptimizers
    #DecisionOptimization