{"id":515,"date":"2007-03-17T14:08:14","date_gmt":"2007-03-17T19:08:14","guid":{"rendered":"http:\/\/UnhinderedByTalent.com\/Phi\/?p=515"},"modified":"2007-03-17T14:20:53","modified_gmt":"2007-03-17T19:20:53","slug":"on-development-and-genetic-diversity","status":"publish","type":"post","link":"https:\/\/UnhinderedByTalent.com\/Phi\/archives\/2007\/03\/17\/on-development-and-genetic-diversity\/","title":{"rendered":"On development and genetic diversity"},"content":{"rendered":"

Screen shot of an EC population using local mating and a simple geography
\nPeeZed<\/a> has a cool column<\/a> in the latest issue of Seed that describes some recent experiments that nicely illustrate how development can both expose and mask genetic variation, thereby allowing or preventing selection from acting on that variation. Very cool stuff, and nicely written.<\/p>\n

It seems that this might be good material to share with the evolutionary computation (EC) crowd. Since there’s rarely any sort of interesting developmental process in EC work, there would be no way for development to expose or mask genetic variation like this. There are some examples in EC that could be seen as examples of this, but not many, and I’ve never really heard someone discuss it in these terms.<\/p>\n

One issue the piece raised for me is the need to have some mechanism to create and maintain genetic diversity in these loci that aren’t under heavy selective pressure. In EC systems (which usually feature haploid genomes, strong selection pressures – engineers want answers! – and panmictic mating), most genetic material tends to fix across the population quite quickly, even if it’s not being selected for. Consequently there’s typically no variation left after even a few generations. This suggests (to me) that if we want these kinds of interesting behaviors in our EC work (and I realize that may be debatable – it depends a lot on what your goals are), then we’re going to have to move past the the simple models that have historically dominated the field.<\/p>\n

The image above is from a paper that Andy Korth and Tyler Hutchison (both UMM ’07) and I just finished yesterday looking at the impact of local mating (which is rarely used in EC) and some sort of “geography” where the fitness function varies across the space the individuals populate. The paper’s pretty rough in places (we ended up being really rushed for various reasons), but the results are nice. The short version is that local mating can definitely help maintain diversity across a population, especially when fitness varies according to one’s location. The image shows different sub-populations forming and persisting as the initial pool of individuals (which started as a little group in the center) spreads out to fill the space. Not earth shattering, but it’s nice to have some quantitative results to support my intuitions on the subject, and it opens up several doors to further work that could be more significant.<\/p>\nNo tag for this post.\n\t

Related posts<\/h4>\n\t