Posted by Shawn Burgess on August 21, 2003 at 09:59:43:
In Reply to: Thoughts about Version 3 posted by David on August 19, 2003 at 07:20:04:
: First of all, Version 3 is not going to happen for a LONG time because I'm fairly optimistic about Version 2. I'd like to let it run for at least a half million votes before we judge it.
: That said, I've been thinking about this issue of distinct genes within a poenome, that is how to allow "good phrases" to become cohesive whole that are unlikely to be broken up by crossover. Snip points would be more likely to occur in between the genes. In the biological world the large regions of junk DNA in between genes accomplish this (although I'm not sure if this is really a fair analogy.)
Dave, the whole crossover concept doesn't really jive with biology and the junk DNA analogy really breaks down here. Crossover typically occurs in a sexually reproducing organism where the DNA content is 2x (diploid) and crossovers ONLY occur on matching strands. Thus there is a mixing up of variants, but all the basic or essential genes (words) are there just in different flavors. This stabilizes the species and makes progressive changes slower. Your poems are obligate haploid or one copy organisms. The evolution of these kinds of organisms (bacteria) is generally very different and also isn't modeled in your program. Crossing over typically does not occur. In your system, you are getting crossover of haploid and once you get a crossover, you have a different organism alltogether. Not wrong, in fact appropriate for the experiment, but not mirroring actual biology. The best way to approach thinking about it (in my mind) is not to try and make it imitate actual evolution, but to generate an system that most efficiently models selective pressure(s) conveying increasing fitness to certain effective combinations. I don't currrently have a good idea as to how to do that unfortunately...