Decision Optimization

Decision Optimization

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


#Analytics
#DecisionOptimization
#DecisionOptimization
 View Only
  • 1.  A question about Integrality Tolerance in Cplex

    Posted 04/26/10 10:13 PM

    Originally posted by: SystemAdmin


    Hi, may I have a question about changing integrality tolerance parameter in cplex?

    In my integer problem, y is the binary variables set( y is 0 or 1). When I used the default integrality tolerance parameter value 1e-5, I got the values for y (-9.38318e-022, 0, 8.28888e-006, 1, 0, 9.05887e-006, 5.42133e-007, 8.51658e-006). But because the coefficients for y is very large, and the round off error is very large, I want to tight the integrality tolerance.

    I used "cplex.setParam(IloCplex::EpRHS, 1e-9);" and "cplex.setParam(IloCplex::EpInt, 1e-9);" in my program to change the parameter values, but the values for y does not change. (I used cplex 11, concert technology)

    My question is why the values for y does not change? If my method is not correct, how can I do it?

    Thank you very much!
    #CPLEXOptimizers
    #DecisionOptimization


  • 2.  Re: A question about Integrality Tolerance in Cplex

    Posted 04/27/10 05:59 AM

    Originally posted by: SystemAdmin


    I guess that the main issue you are facing evolves from bad numerics due to your very large coefficients. Therefore, my primary suggestion is that you try to model your problem in a different way that does not lead to such large coefficients.

    You can also try to set the integrality tolerance to zero. In this case, CPLEX would try to reduce integer infeasibilities as much as possible, but there is no guarantee that it will succeed.
    #CPLEXOptimizers
    #DecisionOptimization


  • 3.  Re: A question about Integrality Tolerance in Cplex

    Posted 04/27/10 04:05 PM

    Originally posted by: SystemAdmin


    Thank you, I will try it.
    #CPLEXOptimizers
    #DecisionOptimization