Which mechanism is respnsible for population size decrease in eaSimple? (deap reference example)

[Andrea BETTATI]

Hi to you all,
I'm diving a bit in the functioning of tha eaSimple algorithm as it is implemented in the regression example. I see that my population size is varying trough generations: it oscillates.

From the docs (see variations) I understood that crossover may be responsible for size augmentation, what I don't get is how the population size may decrease.

To my understanding toolbox.clone() and mutation can not delete indivisuals.

Also, selection in eaSimple is 1:1 replacement since it is called like this

offspring = toolbox.select(population, len(population))

This is really puzzling me, any help is very much appreciated :)

[NITROMEDO]

So did you find an answer?

[François-Michel De Rainville]

I think what you see is the number of modified individuals (evaluated). This is pretty random with probability centred around joint probabilities CXPB and MUTPB.

-- 
You received this message because you are subscribed to the Google Groups "deap-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to deap-users+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/deap-users/34c24c94-1965-430d-ae52-de1a21f1017e%40googlegroups.com.

[Andrea BETTATI]

Hi NITROMEDO,

I think Francois is correct, the overall pop size should be constant during all training.

--