利用分解的行动空间实现医疗保健中高效的离线强化学习（英文版）

英文标题：Leveraging Factored Action Spaces for Efficient Offline Reinforcement Learning in Healthcare中文摘要：本文研究了如何在强化学习的组合行为空间中通过线性 Q 函数分解来更好地处理少见子行动组合的情况，并对该方法进行了理论分析和实验评估，证明了它可以提高数据效率和策略优化的性能。英文摘要：Many reinforcement learning (RL) applications have combinatorial actionspaces, where each action is a composition of sub-actions. A standard RLapproach ignores this inherent factorization structure, resulting in apotential failure to make meaningful inferences about rarely observedsub-action combinations