Deep active inference agents 

using Monte-Carlo methods