TGRL: An Algorithm for Teacher Guided Reinforcement Learning
This addresses the challenge of combining different learning forms in sequential decision-making for AI/robotics applications, representing an incremental improvement over prior heuristic-based methods.
The paper tackles the problem of balancing reinforcement learning and teacher-student learning objectives without relying on heuristics or hyperparameter tuning, by introducing TGRL, which dynamically adjusts teacher supervision based on performance comparisons, resulting in outperforming strong baselines across diverse domains.
Learning from rewards (i.e., reinforcement learning or RL) and learning to imitate a teacher (i.e., teacher-student learning) are two established approaches for solving sequential decision-making problems. To combine the benefits of these different forms of learning, it is common to train a policy to maximize a combination of reinforcement and teacher-student learning objectives. However, without a principled method to balance these objectives, prior work used heuristics and problem-specific hyperparameter searches to balance the two objectives. We present a $\textit{principled}$ approach, along with an approximate implementation for $\textit{dynamically}$ and $\textit{automatically}$ balancing when to follow the teacher and when to use rewards. The main idea is to adjust the importance of teacher supervision by comparing the agent's performance to the counterfactual scenario of the agent learning without teacher supervision and only from rewards. If using teacher supervision improves performance, the importance of teacher supervision is increased and otherwise it is decreased. Our method, $\textit{Teacher Guided Reinforcement Learning}$ (TGRL), outperforms strong baselines across diverse domains without hyper-parameter tuning.