You run N isolated populations that don't cross over and then you only periodically exchange or cross over individuals from the individual populations.
Sometimes gets better results because each isolated ecology, has a probability to just fail and not work; or each can evolve independent partial solutions.
So if there are five computers, may make sense to evolve independently, then cross over or exchange individuals every hundred generations. Maybe dump population or best 5 individuals every epoch or every 100 generations or fixed rate (every N) save each individual to disc every epoch or best 5 individuals or something or the whole population
then do a random cross over later
we will use NAS
so all the servers will have a shared drive
You run N isolated populations that don't cross over and then you only periodically exchange or cross over individuals from the individual populations.
Sometimes gets better results because each isolated ecology, has a probability to just fail and not work; or each can evolve independent partial solutions.
So if there are five computers, may make sense to evolve independently, then cross over or exchange individuals every hundred generations. Maybe dump population or best 5 individuals every epoch or every 100 generations or fixed rate (every N) save each individual to disc every epoch or best 5 individuals or something or the whole population
then do a random cross over later
we will use NAS
so all the servers will have a shared drive