Decision Optimization

Decision Optimization

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


#Analytics
#DecisionOptimization
#DecisionOptimization
 View Only
  • 1.  Limits for Numerical Representation in the CPLEX 12.3 API for MATLAB

    Posted 10/16/11 09:26 PM

    Originally posted by: BerkUstun


    I am currently using the CPLEX 12.3 API for MATLAB 2011b to solve an LP whose entries can take on large numerical values (~10^6 to 10^12 - and hard to 'rescale' these entries). To be clear, the API involves creating and working with Cplex() as opposed to the built-in MATLAB functions.

    I am wondering what the limits are on numerical representation within MATLAB (i.e. what is the largest order of magnitude that I can work with). Also, is a particular number that a variable, objective value or solution can take on before the LP is classified as 'unbounded'?
    #CPLEXOptimizers
    #DecisionOptimization


  • 2.  Re: Limits for Numerical Representation in the CPLEX 12.3 API for MATLAB

    Posted 10/25/11 08:26 AM

    Originally posted by: SystemAdmin


    The callable library that works under the hood of the matlab connector treats any value smaller than -CPX_INFBOUND or larger than CPX_INFBOUND as infinity (CPX_INFBOUND is 1e20).
    Unboundedness of an LP is not declared depending on the order of magnitude of the objective function value. It is declared when the model can be proven to be unbounded, for example because you maximize x and x is not bounded from above.
    If you create problems in which the coefficients have large magnitudes then you should make sure that the range of coefficients is still reasonable, e.g., do not mix 1e12 with 1e-6.
    #CPLEXOptimizers
    #DecisionOptimization


  • 3.  Re: Limits for Numerical Representation in the CPLEX 12.3 API for MATLAB

    Posted 10/25/11 07:22 PM

    Originally posted by: BerkUstun


    Thank you for this!
    #CPLEXOptimizers
    #DecisionOptimization