2026年第59期(总第1203期)

演讲主题：Compact Approximate Linear Programs for Network Revenue Management Under Choice Models

主讲人：张锐 科罗拉多大学博尔德分校利兹商学院副教授

主持人：秦虎 信息管理与数据科学系教授

活动时间：2026年07月11日（周六）09:00-10:30

活动地址：管院大楼105教室

主讲人简介：

张锐是科罗拉多大学博尔德分校利兹商学院战略、创业与运营系的Stone Family特聘学者和副教授。他担任供应链分析硕士项目的学术主任。在此之前，他曾担任运营方向博士项目主任。此外，他还担任 INFORMS Journal on Computing 和 Networks 的副主编。他的研究兴趣包括定量方法，尤其是处方性分析技术。他的研究聚焦于收益管理问题、最后一公里配送，以及社交网络中的影响力最大化问题。他的研究成果发表在 Operations Research、Manufacturing & Service Operations Management、Production and Operations Management、INFORMS Journal on Computing、INFORMS Journal on Optimization、Naval Research Logistics、European Journal of Operational Research 和 Networks 等期刊上。此外，他曾获得多项最佳论文奖。他的一组研究成果入选 2022 年 INFORMS Computing Society（ICS）奖的亚军。

活动简介：

Approximate linear programs (ALPs) have received significant attention in the revenue management literature as a viable method for value function approximation and constructing effective control policies. However, the complexity of the ALPs increases exponentially with problem size, rendering them intractable to solve for practical purposes. To tackle this challenge, several techniques have been proposed, including constraint sampling, column generation, and the formulation of compact ALPs, which no longer grow exponentially with problem size. In this paper, we develop compact ALPs by leveraging an intuitive constraint decomposition method, which allows us to retain the primal forms of their original counterparts. The compact ALPs are versatile enough to accommodate any type of basis function for value function approximation under choice models. To validate the advantage of the new approximation, we conduct a numerical study, demonstrating that we obtain tighter upper bounds and larger expected revenues than those known compact formulations in the literature.