讲座题目：
1.Consumer Choice Probability Models – Temporal Trees, Representation and Identification
2.Transfer Learning, Cross Learning and Co-Learning Across Newsvendor Systems with Operational Data Analytics (ODA)
主讲嘉宾：
1. J. George Shanthikumar
2. Lei Li
时间：2024年7月5日（星期五）上午9:30—11:30
地点：tyc86太阳集团304会议室
欢迎感兴趣的师生参加聆听！
tyc86太阳集团
2024年7月3日
主讲嘉宾简介
1. J. George Shanthikumar is the Richard E. Dauch Chair Professor of Manufacturing and Operations Management and a University Distinguished Professor of Management at the Krannert School of Management, Purdue University, West Lafayette, IN and a Professor Emeritus of Industrial Engineering and Operations Research at the University of California, Berkeley, CA. Before joining Purdue, he was a Chancellor’s Professor of Industrial Engineering and Operations Research at the University of California, Berkeley, CA. He received the B. Sc. degree in Mechanical Engineering from the University of Sri Lanka, Peradeniya, and the M. A. Sc. and Ph. D. degrees in Industrial Engineering from the University of Toronto, Toronto, Canada. He was the president of POMS for the year 2018, is a Fellow of the Institute for Operations Research and Management Science (INFORMS) and Production and Operations Management (POM) Societies.
2. Lei Li is an Assistant Professor in the Department of Logistics and Maritime Studies at The Hong Kong Polytechnic University. He received a Ph.D. degree from Purdue University in 2023, a M.S. degree and a B.S. degree in Management from Zhejiang University, in 2017 and 2015, respectively. His research interests include stochastic modeling, data-integrated modeling, supply chain management, and not-for-profit operations. His research papers have been published in Management Science, Production and Operations Management.
讲座主要内容
1. A consumer choice probability model (CCPM) characterizes the consumer choice probabilities (CCPs) through a set of equality and inequality constraints. We develop a temporal tree consumer choice probability model where a set of branching probabilities fully characterize the consumer choice probabilities of all rational consumer choice models (CCM) such as Random Utility Model (RUM) and Rank List Model (RLM).
We will outline the Temporal Tree Representation of standard CCMs such as Multinomial Logit Model and its variants, Exponomial Model, Markov Chain Model etc. We will show that a suitably defined subclass of Temporal Trees can be uniquely identified for any given rational consumer choice probabilities. This makes the Temporal Tree amenable for Identification using machine learning algorithms with extension to consumer choice models that are built on consumer and/or product attributes.
2. Decision making with limited statistical characterization and limited data is challenging. The typical statistical-machine-learning approaches 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, instead, develop new solution concepts to learn across related systems by adapting the parametric Operational Data Analytics (ODA) framework, which is known to produce uniformly optimal data-integrated decisions in the corresponding parametric settings, for non-parametric decision making. We demonstrate, through the application of newsvendor systems, that transfer learning can, indeed, improve decision performance in the focal system by utilizing a model pre-trained using the ample data in a related system. However, in the lens of the ODA framework, the best transfer-learning decision falls in a subclass of operational statistics, limiting the ultimate optimality. In contrast, the ODA cross-learning approach utilizes the ample data from the related system to mimic the stochastic environment of the focal system. When the data from the old system are sufficiently large, the cross-learning solutions derived outperforms any transfer learning solution, and they are shown to asymptotically approach the parametric ODA solutions. When there are multiple related systems with limited data, we aggregate the data from different systems to create a generic stochastic environment for the decision-making problem, which facilitates the implementation of the parametric ODA solutions. We show that the derived co-learning solutions are asymptotically optimal for the aggregate system and for each sub-system. This approach outperforms the existing data-pooling techniques in the sense that the latter focuses only on the aggregated performance, and the chosen solution may be (asymptotically) suboptimal for individual sub-systems. Our results underscore the roles of domain knowledge and the structural relationships between the data and the decision in designing efficient learning solutions with limited data. Though we demonstrate our development through the application of newsvendor systems, the solutions developed in this study applies to a much wider class of operational decision-making problems that exhibit certain homogenous properties.
当前位置: