Decision Optimization

Decision Optimization

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


#Analytics
#DecisionOptimization
#DecisionOptimization
 View Only
  • 1.  MIP start infeasible

    Posted 08/29/12 09:23 AM

    Originally posted by: JorisK


    Dear,

    Using the Java programming interface of CPLEX, I've built a rather complicated MIP model. The model is initialized with a feasible initial solution by invoking the addMIPStart(vars, values, IloCplex.MIPStartEffort.CheckFeas); function. However, when solving the MIP problem, I get the following error:

    Warning: No solution found from 1 MIP starts.

    which implies that either my initial solution is not feasible, or my model cuts of feasible solutions. In either case I would like to find the constraints which cut of the initial solution. According to the documentation (User's Manual for CPLEX > Refining a conflict in a MIP start) there are multiple approaches to find a minimal conflict set of constraints. I first exported the MIP model to a file using the cplex.exportModel("test.lp"); function. Additionally, I exported the MIP start via the function cplex.writeMIPStart("mipStart.mst"); (see attached documents).

    Finally, I tried to determine the conflict using the Interactive Optimizer as suggested in this thread: http://www.ibm.com/developerworks/forums/thread.jspa?threadID=363875

    CPLEX> read test.lp
    Problem 'test.lp' read.
    Read time = 0.03 sec.
    CPLEX> read mipStart.mst
    MIP start file 'mipStart.mst' read.
    CPLEX> conflict 1

    Refine conflict on 3000 members...

    Iteration Max Members Min Members
    1 0 0

    Minimal conflict: 0 linear constraint(s)
    0 lower bound(s)
    0 upper bound(s)
    Conflict computation time = 0.02 sec. Iterations = 1

    Apparently the approach suggested in the previous thread does not work (anymore) as I do not get any conflicts, or am I doing something wrong?
    #CPLEXOptimizers
    #DecisionOptimization


  • 2.  Re: MIP start infeasible

    Posted 08/29/12 12:27 PM

    Originally posted by: T_O


    Your MIPStart does not satisfy constraint IloC2849 as all its variables are 0. To find this out, I fixed all bounds to your MIPStart and ran the conflict refiner.

    Best regards,
    Thomas
    #CPLEXOptimizers
    #DecisionOptimization


  • 3.  Re: MIP start infeasible

    Posted 08/29/12 06:01 PM

    Originally posted by: JorisK


    Thank you for the effort!
    Out of curiosity: how did you fix all bounds to my MIPStart values so quickly? Is there some function within cplex which can do that, or did you write a script for that? Surely, you did not assign nearly 3000 variables manually :)
    #CPLEXOptimizers
    #DecisionOptimization


  • 4.  Re: MIP start infeasible

    Posted 08/30/12 02:13 AM

    Originally posted by: T_O


    I used search-and-replace in Notepad++ a few times to make the MIPstart look like bounds of an LP-file. Then it was just a simple copy&paste.

    Best regards,
    Thomas
    #CPLEXOptimizers
    #DecisionOptimization


  • 5.  Re: MIP start infeasible

    Posted 08/31/12 12:03 PM

    Originally posted by: SystemAdmin


    On a Linux command line things were even easier:
    sed -n -e 's,^.*name="\([^"]*\)".*value="\([^"]*\)".*$,\1 = \2,p' mipStart.mst
    

    Add the output of that at the end of the "Bounds" section in the LP file and your done.
    #CPLEXOptimizers
    #DecisionOptimization


  • 6.  Re: MIP start infeasible

    Posted 08/31/12 12:01 PM

    Originally posted by: SystemAdmin


    The input sequence to the interactive looks correct.
    I can reproduce the problem here and have filed a bug report. Thank you for reporting this.
    #CPLEXOptimizers
    #DecisionOptimization