Agent's actions as a classification criteria for the state space in a learning from rewards system | Publicación