深度可解释关系强化学习：神经符号方法（英文版）

英文标题：Deep Explainable Relational Reinforcement Learning: A Neuro-Symbolic Approach中文摘要：提出了一种新的框架 Deep Explainable Relational Reinforcement Learning (DERRL)，它结合了神经网络和符号世界的优势来提取可解释的策略。通过在倒计时游戏、积木世界、网格世界和交通等不同环境中的实验证明，DERRL 可以适用于不同的配置和情境，从而具有通用性。英文摘要：Despite numerous successes in Deep Reinforcement Learning (DRL), the learnedpolicies are not interpretable. Moreover, since DRL does not exploit symbolicrelational representations, it has difficulties in coping with structuralchanges in its environmen