News for hackers. Stuff that matters.
Re:Genetic Programming: Evolution of Mona Lisa by randomwalker on Monday December 08, 2008 @06:10PM (#390558)
Dude, pastebin? :-)

Yeah, I guess this is GA alright. Your crossover is not right -- it should "cross over" at one point, not flip flop at each location :-)

[Edit] And there's a lot of in-breeding going on! You want to keep 10-20 sequences from each generation, not just 2.

Re:Genetic Programming: Evolution of Mona Lisa by yters on Tuesday December 09, 2008 @09:55AM (#391318)
Crossing over at one point is not as effective as an even crossover, since it is actually equivalent to an even crossover with one static and one dynamic point. Thus, building blocks in the middle are more likely to be disrupted than building blocks on the ends.

That being said, his approach is also a viable GA operator. It is called a uniform crossover. But really, to qualify as a GA, his solution only needs some kind of variation, mutation, and selection.