My idea is to spawn multiple games and use a master that receives the fitnesses from each instance, calculates a global score and then triggers the training of the networks.
Once that is done it tells the games to reset and run again.
Communication can be done with rpc calls. (what is the fastest implementation?)
My idea is to spawn multiple games and use a master that receives the fitnesses from each instance, calculates a global score and then triggers the training of the networks. Once that is done it tells the games to reset and run again. Communication can be done with rpc calls. (what is the fastest implementation?)