Decision Optimization

Decision Optimization

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

 View Only
  • 1.  Random seed in remote CPLEX instances

    Posted Tue July 14, 2015 11:05 AM

    I have a two part question about the distributed MIP feature in CPLEX 12.6.x. Both pertain to creating multiple remote instances of IloCplex, as in the RemoteParBenders.java example.

    • If I leave the RandomSeed parameter at its default in the remote IloCplex instances, do they all use the default seed for the software release in play, or does CPLEX vary the seeds?
    • Similarly, if I specify a value for RandomSeed by calling setParameter on each instance, do I need to vary the seed argument to avoid all instances using the same seed? (I'm fairly sure the answer to this is "yes", but it can't hurt to check.)

    Basically, what I'm trying to confirm is that if I want to do a racing ramp-up using different seeds but the same parameters otherwise across the remote instances, I'm responsible for setting the seeds in each instance.


    #CPLEXOptimizers
    #DecisionOptimization


  • 2.  Re: Random seed in remote CPLEX instances

    Posted Tue July 14, 2015 12:09 PM

    Originally posted by: Laci Ladanyi


    If you use the remote object directly the you are responsible for setting a different random seed in each of the remote objects. If you use the distmip facilities then by default cplex assigns different random seeds to each worker (of course, you can still override that and assign a seed yourself).


    #CPLEXOptimizers
    #DecisionOptimization


  • 3.  Re: Random seed in remote CPLEX instances

    Posted Tue July 14, 2015 01:09 PM