Useful Policy Invariant Shaping from Arbitrary Advice

Abstract

Reinforcement learning (RL) is a powerful learning paradigm in which agents can learn to maximize sparse and delayed reward signals. Although RL has had many impressive successes in complex domains, learning can take hours, days, or even years of training data. A major challenge of contemporary RL research is to discover how to learn with less data. Previous work has shown that domain information can be successfully used to shape the reward; by adding additional reward information, the agent can learn with much less data. Furthermore, if the reward is constructed from a potential function, the optimal policy is guaranteed to be unaltered. While such potential-based reward shaping (PBRS) holds promise, it is limited by the need for a well-defined potential function. Ideally, we would like to be able to take arbitrary advice from a human or other agent and improve performance without affecting the optimal policy. To achieve this, the current thesis presents a simple method called policy invariant explicit shaping (PIES). We further expose a technical flaw in the recently introduced dynamic potential based advice (DPBA) method and show theoretically and empirically that PIES is a simple alternative that succeeds where DPBA fails.