Academy of Mathematics and Systems Science, CAS
Colloquia & Seminars

Speaker:	徐建宇助理教授,西交利物浦大学
Inviter:	胡庆培教授
Title:	强化学习在工程维修中的应用：结构性质与可行解
Language:	Chinese
Time & Venue:	2023.07.10 19:00-20:00 南楼N402
Abstract:	Reinforcement learning (RL), as one of the three state-of-the-art machine learning areas, is gaining increasing interest in operations management domain. Reinforcement learning models focus on the online decision tasks against some unknown environment, where the agent needs to balance between exploiting historical observations to optimize benefits, and exploring those seemingly sub-optimal choices to improve their knowledge towards the environment. A challenging gap of traditional RL is the notorious computational complexity caused by the curse of dimensionality, which widely exists in operations management (OM) problems where states of systems are evaluated from multiple dimensions. In this work, the author summarizes some important features in OM problems that can greatly release the computational cost of RL approaches, and thus break the barrier of using RL in such problems. Such properties are referred to as structural properties in the work. In addition, it is illustrated in the work that incorporating these structural properties in RL approach leads to some high-level and informative conclusions in the management context.