Data Efficient Methods For Reinforcement Learning