报告题目：Warehouse Assortment Selection with Distribution-Free Guarantees via Birkhoff--von Neumann Decompositions
报告人：林宏远 副教授
邀请人：边博旻 教授
报告时间及地点：2026.10.16 10:00-11:30 经管学院B312
报告人简况：
林宏远，现任中国科学技术大学科技商学院、管理学院特任副教授，博士毕业于新加坡国立大学工业系统工程与管理系。主要研究方向包括收益管理、供应链管理与数据驱动决策，关注电商场景下的定价、产品推荐与仓储运营优化。以第一作者或通讯作者在 M&SOM 和 POM 发表论文。曾获2022年 MSOM 数据驱动研究挑战赛一等奖，以及2026年 CSAMSE 最佳论文奖三等奖。
报告内容摘要：
The rapid growth of e-commerce has intensified the challenge of efficient order fulfillment, as capacity-constrained front-end warehouses cannot carry the full product assortment. An important feature of this problem is that products may be subject to heterogeneous lower and upper bounds on the number of warehouses in which they can be stocked, reflecting requirements such as non-overlap, full coverage, and exact assignment. We study warehouse assortment selection in single- and multi-warehouse systems, where each customer order should be fulfilled entirely from one warehouse. We develop a scalable heuristic based on the Birkhoff--von Neumann (BvN) decomposition. The method solves a linear programming relaxation, converts its fractional stocking decisions into a collection of feasible integral assortments via a Restricted BvN (R-BvN) procedure that generates only a number of candidates linear in the product--warehouse dimension, and returns the best-performing candidate. We establish distribution-free approximation guarantees: the factor is bounded by the maximum order size in a single-front-end-warehouse system and by the maximum order size divided by the positive fulfillment-cost gap between the cheapest and second-cheapest warehouses in a multi-warehouse system.
上一条：公交串车控制：数据驱动学习与机理驱动设计
下一条：Forecasting Volatility with Sentiment: Do we need to catch the hype of Al and LLM?
【关闭】
