My plan is to incorporate probabilistic models in my agents, is it possible to have my agent play for many rounds and save the parameters that it learned ?
Thanks and cheers.
1 Like
No. Your code is always executed from scratch in an empty folder. You can’t store data for furture executions.
That’s kinda disappointing. The games are great, but I was hoping for opportunities to develop AI agents that are more advanced than rule-based agents.
That’s something not currently possible but that we may implement in the future.
You can however retrieve the games and learn with it by going on the battle page and inspect the network (F12 on your browser). That’s hacky and unofficial but that can do the trick ;).
Awesome! Thanks!