Focus Of Attention In Reinforcement Learning