Re: [GP] Anyone knows anything on intra-population effects?
- On Tue, Dec 11, 2012 at 9:43 PM, hielkeopreis <hielkeopreis@...> wrote:
For a project which I am doing, I am planning to build an Asynchronous Cellular EA with Intra-Population Effects. This last bit (the intra-population effects) is what my question is about.
What I mean by intra-population effects, is that the fitness of an individual in a cell, is dependent on the workings of the individuals in the neighbouring cells. This does not mean that it is dependent on the fitness of the neighbouring cells, but only on their working.I would have thought anything with (co)evolving agents or robots would have this going on. Also I'd suggest that in interactive GAs, the user will be influenced by what they have already heard or seen.
In nature, the fitness of an individual is of course dependent on other individuals and family members, but I think these effects are all lumped together as "environment" in population genetics studies.
But otherwise, no I can't help with any pointers - sorry.
Let us know what you find!
So if an individual in a cell changes it's parameters through evolution, its fitness could for example improve. The internal change of the individual could have either a positive, a negative, or no effect at all on the fitness of the individual in the neighbouring cell. Similarly, when the fitness of some individual declines, the fitness of the individual in the neighbouring cell could also improve, decline or have no change at all.
Please note that I am not talking about the propagation effect between neighbours. Individuals in cells can mate with neighbours and through that system of mating, neighbours already have effects on each other. In this case however, I am talking specifically about an existing individual, which would need to be retested if individuals in its neighbourhood would change their parameters.
This has the effect that you will never be finished testing the individuals, because changing one parameter in one individual might change the working on neighbouring individuals. This however, doesn't matter, since it will be a continuously self learning system in production.
And here comes my question: I've read a lot about cellular EAs but I can't find anything on these intra-populational effects. This "intra-population" is of course a word which I thought of myself so I don't even know if I am looking for the correct thing.
Does anybody know anything about these intra-populational effects? Does it exist in literature on EAs or would it be a completely new thing? All tips are welcome!