Any computer scientists in here?

[GenericKen]

Sweet, how fortunate.


Keeping the values atomic makes sense, thanks. I'm working on a custom little thing in flash (yes flash), and I'll post a link if it ever turns out.

Thanks for the help, man!


P.S. Since my GA is going to be time-sensitive, how much of a mathematical improvement do you think the following modification to truncation selection would be?

choose p from pop {
p(k / (1+k)) choose random from population with fitness > pop avg fitness
p(1 / (1+k)) choose random from total population
}

k may be arbitrary, or may be the healthy fitness normalized to the avg fitness:
(avg fitness of pop with fitness>pop avg fitness)/(pop avg fitness)


And I'm pretty sure what you get is a O(n) rough aproximation of roulette wheel selection for n different ps; the roulette wheel implementation is O(n^2) iirc (or maybe O(n logn) if you cumulate the fitnesses and do a search on them). I'd like a second opinion on this, though.