题　目：Transfer Learning, Cross Learning and Co-Learning Across Newsvendor Systems

嘉　宾：Lei Li, Ph.D. Candidate, Purdue University

主持人：花成 助理教授 上海交通大学安泰经济与管理学院

时　间：2022年11月4日（周四）21:00-22:30pm

地　点：腾讯会议（校内师生如需获取会议号和密码，请于11月4日下午17点前发送电邮至managementscience@acem.sjtu.edu.cn）

内容简介:

Decision making with limited data is challenging. A machine learning approach would call for migrating the experience of a related system with ample data through transfer learning or leveraging the similarity of multiple systems with limited data through data pooling. We demonstrate, through the application of newsvendor systems, that transfer learning can significantly improve decision performance in the focal system by leveraging a well-trained solution using the ample data in a related system. However, the best transfer-learned decision is within a subclass of potential solutions, limiting the ultimate optimality in the newsvendor setting. To address this, we propose cross learning by adapting the parametric solutions of Operational Data Analytics (ODA, which is known to generate the uniformly optimal operational statistics) for non-parametric decision making. Under this approach, we utilize the ample data from the related system to mimic the stochastic environment of the focal system, which facilitates the implementation of the ODA solution. The resulting decision significantly improves the performance of the focal system over the transfer-learned solution and is shown to be asymptotically optimal. When there are multiple related systems with limited data, we transform the data from different systems to create a generic stochastic environment for the decision making problem, which facilitates the implementation of the ODA solution. We show that the derived co-learning solution is asymptotically optimal for each involved system, as well as the aggregate system, and outperforms the existing data-pooling techniques, which focus only on aggregated performance. Our results underscore the role of domain knowledge, the statistical similarity among the related systems, and the structural relationships between inputs and outputs, in designing efficient decisions with limited data.

演讲人简介：

Lei Li is a final-year Ph.D. candidate in Supply Chain & Operations Management at Krannert School of Management, Purdue University. Previously, he obtained B.S. in Management (2015) and M.S. in Management Science and Engineering (2017) from Zhejiang University. His fields of interests include supply chain management and notfor-profit operations. His research focuses on the study of dynamic staffing policies and dynamic inventory-pricing policies under supply uncertainty using stochastic modeling, and the study of data-integrated decision models in different domains.

欢迎广大师生参加!