Hacker News Reader

I've been running it for some time now and it seems that the model started to converge towards generation 35 and then diverged after that. i.e. since you consider the loss to be distance from each particle to the food, then the ants ended up somewhat above the food at 35, and then started overshooting it after that. Is that expected behavior?

View on HN

no, if it reaches food, each generation after that should follow the same path.

Thanks for pointing out. I will debug and update :)

However, all ants will not follow the same path but most should.

I did not expect all the ants to follow the same path, but I thought maybe they'll come close since only the best generations are supposed to get passed on with mutations. I'm at generation 810 and the model hasn't improved much. Most ants (except for ~6) first go to the right to avoid the obstacle, and then they travel to the upper-left corner while about 6-8 reach the food.

I hope it would work better with minor tweaks to calculating the loss and selecting new generations :) Good Luck

Reply on news.ycombinator.com