Decision Optimization

Decision Optimization

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

 View Only
  • 1.  Solve MILP

    Posted Mon September 26, 2016 12:11 PM

    Originally posted by: VladimirKrivonozhko


    Hi,

     

    I am trying to solve attached milp problem, but obtain an objective value 9.23470204720299, while it should be 1.0 as other solvers get. I am using NEOS as well as offline apps with default settings. What settings I should use to get correct result?

     

     

    Kind regards.


    #CPLEXOptimizers
    #DecisionOptimization


  • 2.  Re: Solve MILP

    Posted Tue September 27, 2016 04:09 PM

    I ran your problem in the CPLEX 12.6.3 interactive optimizer (default settings). Presolve shrank the model to four constraints and three variables, solved it at the root, and reported an objective value of 1.0. What did you do that resulted in the 9.235 objective value?

    Paul

     


    #CPLEXOptimizers
    #DecisionOptimization


  • 3.  Re: Solve MILP

    Posted Wed September 28, 2016 06:16 AM

    Originally posted by: VladimirKrivonozhko


    Paul,

    Thanks for your reply. Here is the solver's output, that I recieve from NEOS.
     

    Executing on neos-7.neos-server.org

    Welcome to IBM(R) ILOG(R) CPLEX(R) Interactive Optimizer 12.6.2.0
      with Simplex, Mixed Integer & Barrier Optimizers
    5725-A06 5725-A29 5724-Y48 5724-Y49 5724-Y54 5724-Y55 5655-Y21
    Copyright IBM Corp. 1988, 2015. All Rights Reserved.

    Type 'help' for a list of available commands.
    Type 'help' followed by a command name for more
    information on commands.

    CPLEX> New value for default parallel thread count: 4
    CPLEX> Selected objective sense: MINIMIZE
    Selected objective name: R0
    Selected RHS name: RHS
    Selected bound name: BND
    Problem 'cplex.mps' read.
    Read time = 0.00 sec. (0.10 ticks)
    CPLEX> Found incumbent of value 9.234702 after 0.00 sec. (0.01 ticks)
    Tried aggregator 1 time.
    MIP Presolve eliminated 3 rows and 197 columns.
    MIP Presolve modified 13 coefficients.
    Reduced MIP has 4 rows, 3 columns, and 11 nonzeros.
    Reduced MIP has 2 binaries, 0 generals, 0 SOSs, and 0 indicators.
    Presolve time = 0.00 sec. (4.22 ticks)
    Probing fixed 0 vars, tightened 1 bounds.
    Probing time = 0.00 sec. (0.00 ticks)
    Presolve time = 0.00 sec. (0.01 ticks)

    Root node processing (before b&c):
      Real time = 0.00 sec. (4.25 ticks)
    Parallel b&c, 4 threads:
      Real time = 0.00 sec. (0.00 ticks)
      Sync time (average) = 0.00 sec.
      Wait time (average) = 0.00 sec.
                              ------------
    Total (root+branch&cut) = 0.00 sec. (4.25 ticks)

    Solution pool: 1 solution saved.

    MIP - Integer optimal solution: Objective = 9.2347020472e+00
    Solution time = 0.00 sec. Iterations = 0 Nodes = 0
    Deterministic time = 4.25 ticks (1072.65 ticks/sec)

    CPLEX>

     

    My Interactive Optimizer 12.6.0.0 give the same result. Is there any possibility of solving this problem using this version of optimizer?

     


    #CPLEXOptimizers
    #DecisionOptimization


  • 4.  Re: Solve MILP

    Posted Wed September 28, 2016 09:37 AM

    Vladimir,

    Your output matches the bulk of what CPLEX 12.6.3 gives me (other than small differences in solution times). The catch is that 12.6.3 also does some probing, which apparently the version on NEOS does not. If I turn off probing, 12.6.3 gives the same incorrect result that you get.

    CPLEX never gets to branching on your problem: it appears to declare victory after the presolve stage. If I turn presolve off (along with turning probing off, to mimic 12.6.0), 12.6.3 gets the correct solution, albeit more slowly than with default settings. Leaving presolve on but turning off repeatpresolve also seems to work.

    There are a large number of presolve settings, and there may be some other setting that also will solve the problem, at least with 12.6.3. I suggest turning off repeatpresolve on 12.6.0 as your first attempt -- unless, of course, you can upgrade CPLEX to 12.6.3.

    Paul

     


    #CPLEXOptimizers
    #DecisionOptimization


  • 5.  Re: Solve MILP

    Posted Thu September 29, 2016 05:35 AM

    Originally posted by: VladimirKrivonozhko


    Paul,

    Thank you very much! Turning off repeatpresolve solves my problem.

     

    Vladimir


    #CPLEXOptimizers
    #DecisionOptimization