Originally posted by: SystemAdmin

I am solving a MIP problem usign CPLEX c++ library
I set no cuts, no heuristics and no presolve
After root node, the algorithm spend so much time doing something that i can understand (before branching).
Could you please tell me what the algorithm does?
Thanks in advance
Paula
Nodes Cuts/
Node Left Objective IInf Best Integer Best Node ItCnt Gap

0 0 15.0000 231 15.0000 302
Presolve time = 0.00 sec.
Initializing dual steep norms . . .

Iteration log . . .
Iteration: 1 Dual objective = -13105.435624
Iteration: 121 Dual objective = -13089.303420
Iteration: 213 Dual objective = -12085.177051
Removing shift (1).

Iteration log . . .
Iteration: 1 Dual objective = -11581.043900
Using devex.

Iteration log . . .
Iteration: 1 Objective = -6278.676060
Iteration: 35 Objective = -9588.232445
Iteration: 132 Objective = -15677.106722

Iteration log . . .
Iteration: 1 Dual objective = -17780.016487
Iteration: 7 Dual objective = -17778.840796
Using devex.

Iteration log . . .
Iteration: 1 Objective = -13966.388552
Using devex.

Iteration log . . .
Iteration: 1 Objective = -15664.124769
Using devex.

Iteration log . . .
Iteration: 1 Objective = -16837.188128
Iteration: 5 Objective = -17669.193761
Using devex.

Iteration log . . .
Iteration: 1 Objective = -16507.675023
Using devex.

Iteration log . . .
Iteration: 1 Objective = -16840.010392
Using devex.

Iteration log . . .
Iteration: 1 Objective = -16507.674434

Iteration log . . .
Iteration: 1 Dual objective = -16184.249651
Iteration: 4 Dual objective = -16174.982929
Removing shift (1).
Iteration: 16 Dual infeasibility = 0.000003
Iteration: 22 Dual objective = -16170.249542
Using devex.

Iteration log . . .
Iteration: 1 Objective = -17177.522432
Using devex.

Iteration log . . .
Iteration: 1 Objective = -16511.029461
Using devex.

Iteration log . . .
Iteration: 1 Objective = -16843.787051
Using devex.

Iteration log . . .
Iteration: 1 Objective = -17177.522432
Using devex.

Iteration log . . .
Iteration: 1 Dual objective = 15.000000

Nodes Cuts/
Node Left Objective IInf Best Integer Best Node ItCnt Gap

0 0 15.0000 8 Cuts: 168 303

Iteration log . . .
Iteration: 1 Dual objective = 15.000000
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

I think the iterations you see come from initialization of strong branching.
You can disable strong branching by changing parameter IloCplex::VarSel.
Or you can limit the work spent on strong branch by changing parameters IloCplex::StrongCandLim and IloCplex::StrongItLim.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

Thanks for your answer!
I think that this is not the problem since I set CPX_VARSEL_PSEUDOREDUCED.
I tried CPX_VARSEL_STRONG with StrongCandLim = 1 and StrongItLim=1 and the behaviour is the same.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

OK, could you please use IloCplex::writeParam() to write out a file with changed parameters right before doing IloCplex::solve() and attach that file here?
If possible, could you also attach the model?
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

Param File
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

OK, the situation is as follows: After the relaxation is solved CPLEX tries to do some special things to reduce the number of integer infeasible variables (sorry, I cannot go into more details). These special things are based on solving LPs and that is where the output you see comes from.
You can disable these special strategies by using a hidden parameter:

cplex.setParam(IloCplex::IntParam(2132), -1);

Caveat:

• Hidden parameters are not part of the official public API. They may change or even get removed between releases without notice.
• Disabling the strategies to reduce the number of integer infeasible variables may affect performance in a bad way.

#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

Model
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

Output
#CPLEXOptimizers
#DecisionOptimization