Reward Estimation for Variance Reduction in Deep Reinforcement Learning - View it on GitHub