Special considerations when using ActorCritic MDPSolver

[zju...@tenkiv.org]

I have started experimenting with Burlap and after trying the tutorials I tried to learn more by switching out the MDPSolver for the various examples. I could get most of them to work but I am having quite a bit of difficulty with ActorCritic. Are there some special considerations that need to be taken with ActorCritic? Or, even better, is there a tutorial / example that uses ActorCritic? After looking through the source code it seems like the critic updates the actors policy automatically and I shouldn't need to do anything externally. Sorry for the basic question, I am just getting started with Burlap and reinforcement learning in general.