Policy Transfer using Reward Shaping
Tim Brys, Anna Harutyunyan, Matthew Edmund Taylor, Ann Nowé · 2015
Transfer learning has proven to be a wildly successful ap-proach for speeding up reinforcement learning. Techniques often use low-level information obtained in the source task to achieve successful transfer in the target task. Yet, a most general transfer approach can only assume access to the out-put of the learning algorithm in the source task, i.e. the learned policy, enabling transfer irrespective of the learning algorithm used in the source task. We advance the state-of-the-art by using a reward shaping approach to policy trans-fer. One of the advantages in following such an approach, is that it firmly grounds policy transfer in an actively develop-ing body of theoretical research on reward shaping. Exper-iments in Mountain Car, Cart Pole and Mario demonstrate the practical usefulness of the approach.