Then I am afraid the hacks I had in mind won't work.  You will have to write your own crossover pipeline.  My suggestion is for you to override prepareToProduce to pre-generate an array of random crossover points equal to the number of individuals in a subpopulation (call it "crossovers"), and to set a variable called sub=-1  and another variable called count=0 Then during produce() you do something like this:

	if (sub < subpopulation)
		count = 0;  // reset		

then you use crossovers[count] to determine your pivot location

Then finally you say



On Dec 10, 2014, at 4:22 AM, Nikola <[log in to unmask]> wrote:

> Hi Sean,
> in subpopulations different individuals are selected and subpopulations have different fitness.
> Nikola