Posted by Phil M on March 19, 2004 at 03:32:59:
In Reply to: evolving evolution mechanisms posted by peter on March 18, 2004 at 18:59:40:
: For the next starting group of poems maby instead of just a new evolution sceme several different variable ways of breeding poems should be used and the more successful ones be used more often. This is sorta like how humans evolved sex a better form of repreduction in many cases than say asexual reperduction. just a thought
The way I first saw that happening is having separate 'worlds', each with their own breeding rules (e.g. rate of mutation, limits of breeding capability/death, etc, and proportions of sexual/asexual/trisexual production) where a proportion of poem comparisons is between members of the same world (giving poem survival advantage to the victor as per this single system) and the corresponding proportion being between representatives of two worlds (either 'champions' or random) to give the entire /world/ a survivial advantage. Then the parameters (and possibly populations) of worlds are mixed'n'matched to create new worlds.
That meta-evolution got me thinking about meta-evolution (the way in which the worlds are chosen) and meta-meta-evolution (how the universe of worlds of poems are distributed), and then I realised I was being silly.
So, instead of that, rather than enforcing two-partner copulation, how about tagging a breeding preference so that one species might tend to evolve a very limited same-species policy of breeding, some poems might tend to go icky about anything other than asexual spawning but with a high mutation rate and some poems (that may gravitate to their own species) gather a hareem of like-minded poems for three-, four- or five-way interbreeding of some kind...
There'd need to be the code available to allow all those options to happen, indepentantly, and work out what happens when a polygamist tries to include an asexually-prefering partner into the mix (doubtless there'd be a chance of that, like there's cross-species copulation at the moment). Would the mutation rates be shared out among the offspring 'as is', I wonder (like species is at the moment) or averaged out and shifted randomly for each. In the end, though, you should find that anything that approaches either 0% or 100% word mutation would be selected against in both asexual and sexual reproduction (the former being more bearable for where recombination occurs as well, but 100% effectively being a completely new poem with no relation to the parent(s) or the characteristcs were advantagously exhibited in the previous generation).
Sorry, a lot of food for thought there. I'm pretty sure we're going way beyond the practical limits of this system, though it may well be applicable for one in which a lot more time and energy is put into breeding (such as automated GP working with a standard 'fitness function', rather than our own fickle selves in this more manual process...)