Action selection for composable modular deep reinforcement learning

• المؤلفون الرئيسيون: GUPTA, Vaibhav, ANAND, Daksh, PARUCHURI, Praveen, KUMAR, Akshat
• التنسيق: text
• اللغة: English
• منشور في: Institutional Knowledge at Singapore Management University 2021
• الموضوعات: Reinforcement learning; Coordination and control; Deep learning; Databases and Information Systems
• الوصول للمادة أونلاين: https://ink.library.smu.edu.sg/sis_research/6900; https://ink.library.smu.edu.sg/context/sis_research/article/7903/viewcontent/Action_Selection_for_Composable_Modular_Deep_Reinforcement_Learning.pdf

In modular reinforcement learning (MRL), a complex decision making problem is decomposed into multiple simpler subproblems each solved by a separate module. Often, these subproblems have conflicting goals, and incomparable reward scales. A composable decision making architecture requires that even t...