Print

Print


The eval.subpop.i.* params are for evaluation purposes only.  If you  
write a new evaluator, you would not need them.

On Jan 10, 2007, at 8:16 AM, SUBSCRIBE ECJ-INTEREST-L T. S. Yo wrote:

> Hi,
>
> I have a question about coevolution, to be more specific, it is  
> about the
> class MultiPopCoevolutionaryEvaluator.
>
> In this class, a set of parameters with prefix "eval.subpop.i" are  
> defined
> for each subpopulation, i.e., "num-elites", "num-random-ind", "num- 
> ind", and
> "select".
>
> Because I am writing a new Evaluator class for coevolution, I have the
> following  question : is it "necessary" to use these parameters to  
> define
> the breeding of each subpopulation in coevolution?  May I just  
> define two
> breeding pipeline like this:
> ----------------------------------------------
> pop.subpops        = 2
>
> breed.elite.0    = 10
> pop.subpop.0.species.pipe  = ec.vector.breed.VectorMutationPipeline
> pop.subpop.0.species.pipe.source.0 =  
> ec.vector.breed.VectorCrossoverPipeline
> pop.subpop.0.species.pipe.source.0.source.0 = exp.util.RankSelection
> pop.subpop.0.species.pipe.source.0.source.1 = exp.util.RankSelection
>
> breed.elite.1    = 10
> pop.subpop.1.species.pipe  = ec.vector.breed.VectorMutationPipeline
> pop.subpop.1.species.pipe.source.0 =  
> ec.vector.breed.VectorCrossoverPipeline
> pop.subpop.1.species.pipe.source.0.source.0 = exp.util.RankSelection
> pop.subpop.1.species.pipe.source.0.source.1 = exp.util.RankSelection
> -----------------------------------------------
> and ignore these eval.subpop.i.* in my own Evaluator class?
>
> Actually, in the class "CompetitiveEvaluator", there is no such  
> parameters,
> so I wonder if I must handle the selection in an Evaluator class.
>
> Thanks in advance.
>
> Ting-Shuo Yo