3 ms·
Certainly a two stage fitness process would work, or as mentioned using additional controls on the simulation to avoid idiosyncrasies. It isn't that hard to bui
by ChuckMcM 11y ago
Certainly a two stage fitness process would work, or as mentioned using additional controls on the simulation to avoid idiosyncrasies. It isn't that hard to build a small genetic 'evolver' (my favorites are some of the 'walking' ones which race blob things). But you can evolve all sorts of complex things if you are willing to wait long enough and don't care if you understand the result or not. I built one to evolve a response packet to ssh clients to kill them (I got tired of botnets trying to brute force passwords on a server I ran for a while), all you need is a fitness function, a mutate function, and time :-).