光华讲坛——社ng28南宫国际app名流与企业家论坛第6943期
主题：服务系统中兼顾效率和公平的资源分配方法研究
主讲人：香港科技大学商南宫28加拿大软件 吕国栋助理教授
主持人：工商管理南宫28加拿大软件 章宇教授
时间：6月2日10:00-11:00
地点：柳林校区诚正楼1122
主办单位：工商管理南宫28加拿大软件 科研处
主讲人简介：
Dr. Guodong Lyu is an Assistant Professor at The Hong Kong University of Science and Technology. He was named the Star Faculty of HKUST in 2024 and received the HKUST Faculty Recognition Award in 2025. He also received the NSFC Excellent Young Scientist Scheme in 2024. His research focuses on data-driven decision-making with applications in supply chains, urban transportation, logistics, and public-sector operations. Methodologically, he works on online decision-making, stochastic programming, and distributionally robust optimization. His research has been published in journals including Management Science, Operations Research, and Manufacturing & Service Operations Management. His research has also been recognized through several paper awards, including Finalist in the INFORMS George B. Dantzig Dissertation Award Competition, Outstanding Paper Award from the Urban SIG of the INFORMS TSL Society, and Finalist in the POMS SCM College and Service Operations College Best Student Paper Competition.
吕国栋博士是香港科技大学的助理教授。他于2024年被评为港科大杰出教师，并于2025年获得港科大教师表彰奖。同年，他还获得了国家自然科学基金优秀青年科学基金项目资助。他的研究主要聚焦于数据驱动的决策方法，应用于供应链、城市交通、物流以及公共部门运营等领域。在方法论方面，他主要研究在线决策、随机规划以及分布鲁棒优化。他的研究成果发表在Management Science、Operations Research以及Manufacturing & Service Operations Management等国际顶级期刊上。此外，他的研究还获得了多项论文奖项，包括：INFORMS乔治·丹齐克博士论文奖最终入围奖、INFORMS运输科学学ng28南宫国际app城市特别兴趣小组杰出论文奖，以及生产与运营管理学ng28南宫国际app供应链管理与服务运营两个专业方向的最佳学生论文奖最终入围奖。
内容提要：
Balancing efficiency and fairness in allocation design is a fundamental challenge. The Shapley value, operationalized via the random arrival rule, ensures expected fairness but is not designed to optimize efficiency. Weighted Shapley values introduce exogenous weights, yet still lack a systematic mechanism for efficiency improvement. We propose a new allocation rule that samples weight vectors from a carefully constructed distribution, where each vector governs the trade-off between global efficiency and individual agents. Aggregating across these weighted allocations preserves the fairness guarantees of Shapley allocations while systematically enhancing efficiency and ensuring robustness across scenarios. A key innovation is that the weight distributions are designed via online convex optimization, transforming OCO into a principled tool for fairness-efficiency trade-offs and enabling scalable computation in large stochastic settings.
在分配设计中平衡效率与公平性是一个根本性的挑战。Shapley值通过随机到达规则实现，确保了预期的公平性，但并非为了优化效率。加权Shapley值引入了外生权重，但仍缺乏系统性的效率提升机制。我们提出了一种新的分配规则，从精心构建的分布中采样权重向量，每个向量控制全局效率与个别代理之间的权衡。在这些加权配置之间进行聚合，既能保证Shapley配置的公平性，又能系统地提升效率，确保各场景的稳健性。一项关键创新是权重分布通过在线凸优化设计，使OCO成为公平与效率权衡的原则性工具，并实现大规模随机环境中的可扩展计算。
En
