[Icfp04-discuss] Evolutionary Ants

[Detlef Pleiss]

> What are good criteria for good ants? Collecting much food or collecting
> more than the other? I used the second and saw, that the collecting rate
> decrease sometimes from 300 down to 10 during the evolution. Dont know
> why that could be better....

if the others at that point collect even less, than it's better.
I used the mean value of collected food over several games in different
worlds as to avoid random hits or world spezialization.
A new ant had to yield a better mean food/game value over at least half the
number of games that the best-so-far program played in order be become the
new best-of-breed. I introduced that "half the number" later, I first
required the same number of games, but that was too much as each new
generation was required to win after twice the number of games.

ciao - Det.