The MIT notes -especially the bit about Lamarckian evolution – got me thinking about something I’d done a while ago. For my Masters, I wrote a paper on a new style of genetic algorithm, modelled not on the way genes evolve, but on the way human groups evolve.
Anyway, if you’re interested, I’ve put up a fuller outline and some experiment results here.
My basic idea was this:
– genes/chromosomes form the metaphor for genetic algorithms, and these are ‘bred’ in the computer to come up with solutions to problems.
– because they follow Darwinian metaphor they are random and unpurposive.
– human groups (on the face of it at least) evolve differently.
– they hire, fire, court martial, exile, worship, cannonise, that sort of thing. They are at some level ordered and purposive.
– so I thought it would be interesting to develop a “GA” based on a human groups metaphor, where instead of genes you have people, each with randomly assigned “traits”, who are semi-intelligently chosen for work within evolving groups.
Last time I posted this I got a tirade of abuse for bastardising Darwin, so to try to avoid that can I say now that no, I do not think Darwin was a fool, and no I do not think humans are divorced from nature and the forces that she brings. It was just an idea for a (possibly) quicker algorithm. Maybe it could be “remetaphored” – not people but memes?
I should also mention that a certain Matthew Stanfield helped (but no more!) with the acronym.
Links:
The Evolved Group Generator Algorithm (EGG) – http://174.132.139.176/~monkeyma