Explainable Reinforcement Learning via Model Transforms Mira Finkelstein

Understanding emerging behaviors of reinforcement learning (RL) agents may be difficult since such agents are often trained in complex environments using highly complex decision making procedures.