香港中文大学（深圳）数据科学学院Guillermo Gallego教授、高品教授、王子卓教授、叶荫宇教授、张经纬教授，以及他们团队中的三位博士生和一位本科生，共六篇论文被 2024 年国际互联网经济学术会议（Conference on Web and Internet Economics，WINE）录用。

WINE 是计算经济学领域的重要国际学术会议之一，也是计算经济学领域中首个被中国计算机学会（CCF）推荐为 A 类的国际会议。

WINE 简介
 

在过去的十余年中，理论计算机科学、人工智能、运筹学和经济学的研究共同解决了与激励和计算相关的诸多问题。这一类问题共同构成了计算经济学的研究对象。计算经济学研究的问题在涉及到大量不同人群的网络和互联网等应用领域尤为重要。也因此，十几年来，计算经济学这一领域越来越受到学术界的重视，并取得了长足的发展。

国际互联网经济学术会议（Conference on Web and Internet Economics，WINE）是计算经济学领域的国际顶级学术会议之一。首届WINE由邓小铁教授和叶荫宇教授于2005年在香港共同发起。此后，WINE在包括北京大学、牛津大学、哈佛大学、斯坦福大学等亚、欧、美三大洲的国际知名学府循环举办，每年举办一次，2024年是其第20届会议。WINE 2024 将于 2024 年 12 月 2 日至 5 日在英国爱丁堡举行，由爱丁堡大学信息学院主办。

来源：CCF计算经济学专业组官方公众号、WINE 2024官网

学生作者简介 
*按姓名首字母排序

博士生

李国凯
2020级博士生
港中大（深圳）数据科学学院
数据科学博士
指导老师：王子卓、高品

刘逸诚
2020级博士生
港中大（深圳）数据科学学院
数据科学博士
指导老师：王子卓、高品

杨宗森
2022级博士生
港中大（深圳）数据科学学院
数据科学博士
指导老师：王子卓、高品

本科生 

兰海翔
2019级本科生
港中大（深圳）理工学院
数学与应用数学专业
现于哥伦比亚大学运筹与系统工程系攻读博士

指导教授简介 
*按姓名首字母排序

Guillermo Gallego
校长学勤讲座教授
运筹学学科负责人
康奈尔大学博士

INFORMS会士、INFORMS MSOM杰出会士，HKIE会士，曾获INFORMS 2011收益管理与定价领域历史奖、2012 实践奖、2016 影响力奖，曾任哥伦比亚大学工业工程与运筹学系主任和香港科技大学工业工程与决策分析系主任、全球Top 2%顶尖科学家

研究领域：动态定价、离散选择模型、品类优化、定价分析、动态规划

个人简介：Guillermo Gallego曾任哥伦比亚大学刘氏家族冠名教授、香港科技大学嘉柏有限公司冠名工程学教授、国际运筹与管理协会会士（2012年）、制造与服务运营管理协会杰出会士（2013年）、中国香港工程师学会会士（2016年），也是被公认为现代动态定价先驱之一的国际学者。他曾获2011年INFORMS收益管理与定价领域历史奖、2012年INFORMS实践奖、2016年INFORMS影响力奖、2005年和2021年INFORMS收益管理与定价领域奖。此外，他是唯一一位同时获得《管理科学》和《运筹学》这两个领域顶级期刊的最佳论文奖的学者。Gallego教授曾任哥伦比亚大学工业工程与运筹学系主任（2002-2008年）和香港科技大学工业工程与决策分析系主任（2016-2022年）。Gallego教授一直担任博士生导师，将学生们送往全球顶尖的机构，包括中国的复旦大学、香港中文大学（深圳）和上海财经大学；美国的斯坦福大学、密歇根大学、约翰霍普金斯大学和德克萨斯大学达拉斯分校；加拿大的多伦多大学和麦吉尔大学，以及英国的伦敦经济学院。

高品
助理教授
香港科技大学博士

曾在顶刊Management Science与Operations Research发表论文，曾任金融科技公司定量分析师

研究领域：收益管理、推荐系统、算法设计、机制设计和社交媒体中的应用

个人简介：高品教授现为香港中文大学（深圳）助理教授。高教授于2021年获得香港科技大学的博士学位。在此之前，他在业界待了两年，并分别于2013年和2015年在武汉大学和香港科技大学获得物理学学士和硕士学位。高教授目前的研究兴趣包括收益管理、推荐系统、算法设计和机制设计。

叶荫宇
教授（特聘）
斯坦福大学博士

2006年INFORMS Farkas奖（首届获奖者）、2009年约翰·冯·诺依曼理论奖、2012年国际数学规划大会（ISMP）Tseng Lectureship 奖（首届获奖者）、2014年美国应用数学学会（SIAM）优化奖

研究领域：连续和离散优化、数据科学及应用、数字算法设计及分析、算法博弈及市场均衡、运筹及管理科学

个人简介：叶荫宇曾任斯坦福大学管理科学与工程系及计算数学工程研究院李国鼎讲席教授。他的主要研究方向为连续和离散优化、数据科学及应用、数字算法设计及分析、算法博弈及市场均衡、运筹及管理科学等；他和其他科学家开创了内点优化算法、锥规划模型、分布式鲁棒优化、在线线性规划和学习、强化学习和马可夫过程算法分析等。他多次获得科学奖项：包括包括2006年因在最优化领域做出的基础性贡献而获得的INFORMS Farkas奖（首届获奖者）、2009年因在运筹学和管理科学领域做出的根本性持续贡献而获得的约翰·冯·诺依曼理论奖、国际数学规划2012 Tseng Lectureship Prize（每三年颁发一次）、2014美国应用数学学会优化奖（每三年颁发一次）等。根据谷歌学术统计，目前他的文章被引用总计超过59000次。

王子卓
教授
副院长（教学）
斯坦福大学博士

杉数科技创始人、首席技术官，曾获国家自然科学基金原创探索计划项目资助，Adobe数字营销研究奖、CSAMSE最佳论文奖，原明尼苏达大学副教授

研究领域：随机和鲁棒优化、数据驱动决策问题、定价和收益管理

个人简介：王子卓博士现为数据科学学院教授、副院长（教学）。王子卓教授于2007年本科毕业于清华大学数学与应用数学系，2011年获得斯坦福大学金融数学硕士学位，2012年获斯坦福大学管理科学与工程博士学位。王子卓曾任职明尼苏达大学工业与系统工程系助理教授、副教授。

王子卓教授的主要研究方向为在线机器学习及收益与运营管理。在机器学习方面，王子卓教授在在线学习方面做了开创性的工作，对在线线性规划、在线凸规划问题中获得了开创性的结果。在收益管理方面，王子卓教授对消费者行为，商品定价和市场量化营销有着深入研究。他在运筹学和管理科学国际顶尖杂志上发表过超过50篇文章，在国内国际会议上多次应邀进行报告，并担任Management Science，Operations Research，M&SOM, POMS等顶级管理科学杂志编委，并且获得多项学术奖项。王子卓教授曾经或正在主持包括来自中国国家自然科学基金、美国国家自然基金等多项研究项目，总金额超千万元人民币。

王子卓教授在工业界有着丰富的经验，曾参与IBM定价项目，也曾为希捷、美国运通等做过项目咨询，也曾在华尔街量化基金担任过研究员。2016年起，王子卓与他人共同创立杉数科技并担任CTO，过去五年在国内为超百家企业做智能决策方面的咨询与服务，客户包括京东，顺丰，滴滴，华为，南航等国内领头企业。

张经纬
助理教授
杜克大学博士

国际著名期刊Management Science、INFORMS Journal on Computing审稿人，曾任哥伦比亚大学博士后

研究领域：近似动态规划、随机模型

个人简介：张经纬教授现任香港中文大学（深圳）数据科学学院助理教授。张教授于2018年获得新加坡国立大学理学及社会科学学士学位，之后获得杜克大学决策科学博士学位。张教授的研究兴趣包括近似动态规划的理论及应用。

论文介绍 

1. LP-based Control for Network Revenue Management under Markovian Demands
作者：Haixiang Lan, Guillermo Gallego, Zizhuo Wang and Yinyu Ye

论文摘要：We study a network revenue management (NRM) problem, where the seller irrevocably accepts or rejects customers' requests to maximize the total revenue over a finite time horizon subject to limited resources. We consider the setting where customer arrivals are dependent and follow the Markov chain choice model. We propose a deterministic linear program (DLP) which serves an upper bound of this problem, and based on which randomized allocation policies are designed. Specifically, when the customer's arrival distribution is known, we design a static probabilistic allocation (SPA) policy and show that the regret, measured by the difference between its expected revenue and the DLP's optimal value, scales as the square root of the time horizon. When customer's arrival distribution is unknown, we design an online probabilistic allocation (OPA) policy by sequentially learning the parameters, achieving a similar regret under mild conditions. We show that the asymptotic regrets of our algorithms are optimal in the respective settings. Furthermore, we conduct numerical experiments which validate the superior performance of the proposed algorithms.

2. Simultaneous vs Sequential: Optimal Assortment Recommendation in Multi-Channel Retailing
作者：Yicheng Liu, Xiao Alison Chen, Yan Liu, Zizhuo Wang

论文摘要：We study a multi-store assortment planning problem in which a seller centrally decides the assortment of products in each store. Each store is visited by a certain fraction of customers. Upon arrival, customers observe the products offered in the current store. The seller can either 1) show products offered in other stores to customers simultaneously (called simultaneous offering strategy) or 2) show products offered in other stores to customers if they decide not to purchase in the current store (called sequential offering strategy). Customers may incur a disutility if they choose products from other stores. We study the optimal assortment planning problem under each of the two strategies and compare their performance. We show that under the simultaneous strategy, the optimal assortment for each store is revenue-ordered, and the seller cannot do better than if it operates each store separately. In contrast, adopting the sequential strategy can significantly increase the seller's revenue. In particular, when customers' disutilities of purchasing from other stores are universally homogeneous, the optimal assortment under the sequential offering strategy is revenue-ordered for each store, and the store with a lower (higher, resp.) demand should offer a larger (smaller, resp.) assortment to facilitate the sequential selling process. We then compare the two offering strategies in terms of the seller's revenue and customer surplus and analyze the impact of parameters on the revenue comparison. We also consider several extensions, e.g., heterogeneous valuation, capacity constraint, and partial recommendation, and find that our main results still hold qualitatively. Finally, we study the joint assortment and pricing problem under both strategies. The sequential offering strategy may be a better strategy for sellers in practice. We provide detailed guidelines for sellers to implement the sequential strategy, such as regarding the optimal assortment decisions and when sellers can exploit the most benefits.

链接：https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4592172

3. Infrequent Resolving Algorithm for Online Linear Programming
作者：Guokai Li, Zizhuo Wang, Jingwei Zhang

论文摘要：Online linear programming (OLP) has gained significant attention from both researchers and practitioners due to its extensive applications, such as online auction, network revenue management and advertising. Existing OLP algorithms fall into two categories: LP-based algorithms and LP-free algorithms. The former one typically guarantees better performance, even offering a constant regret, but requires solving a large number of LPs, which could be computationally expensive. In contrast, LP-free algorithm only requires first-order computations but induces a worse performance, lacking a constant regret bound. In this work, we bridge the gap between these two extremes by proposing an algorithm that achieves a constant regret while solving LPs only O(loglogT) times over the time horizon T. Moreover, when we are allowed to solve LPs only M times, we propose an algorithm that can guarantee an O(T^{(1/2+ϵ)^M}) regret. Furthermore, when the arrival probabilities are known at the beginning, our algorithm can guarantee a constant regret by solving LPs O(loglogT) times, and an O(T^{(1/2+ϵ)^(M-1)}) regret by solving LPs only M times. Numerical experiments are conducted to demonstrate the efficiency of the proposed algorithms.

链接：https://arxiv.org/abs/2408.00465

4. Robust Optimal Selling Mechanism under Non-linear Utility
作者：Pin Gao, Yicheng Liu, Shixin Wang, Zhen Wang, Zizhuo Wang

论文摘要：We consider a robust mechanism design problem in which a seller wishes to choose a selling mechanism to maximize the revenue obtained from a buyer with private valuation information and a nonlinear utility function. The objective is to provide a profit guarantee for the seller across all possible valuation distributions whose mean and variance are consistent with the prior information. When the utility function is of a quadratic form and the coefficient of variation (i.e., the ratio between the standard deviation and the mean) is low, we find that a quadratic pricing mechanism yields the optimal revenue guarantee. In this case, the worst-case valuation distribution follows a generalized Pareto distribution. However, when the coefficient of variation is high, a unit price mechanism is optimal. We further extend our analysis to general elasticity utility functions and investigate how the optimal worst-case revenue varies with elasticity and coefficient of variation. Furthermore, given the widespread implementation of a two-part tariff mechanism in real-world scenarios, we also study the optimal two-part tariff mechanism. We show that under the quadratic utility form, the optimal two-part tariff pricing strategy achieves at least 80% of the optimal worst-case revenue when compared with the optimal robust mechanism. We also develop corresponding performance guarantees when the elasticity coefficient is less than two. However, when the elasticity coefficient exceeds two, the competitive ratio of the two-part pricing scheme goes to zero as the variance of valuation grows to infinity.

链接：https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4731184

5. When to Push Ads: Optimal Mobile Ad Campaign Strategy under Markov Customer Dynamics
作者：Guokai Li, Pin Gao, Zizhuo Wang

论文摘要：We investigate a seller’s optimal advertising campaign strategy targeting customers who interact with the seller over time. We model customers’ engagement as a continuous-time Markov chain with two states, active and inactive. While in the active state, customers make purchases according to a Poisson process, with each purchase yielding a specific reward; in contrast, customers in the inactive state make no purchases. The seller can use advertisements to activate customers and the activation probabilities may be affected by customers’ fatigue to promotion. The objective of the seller is to maximize the long-term expected average profit by designing an optimal ad campaign strategy based on customers’ purchase histories. For a single customer without budget constraints, we demonstrate the optimality of a triple-threshold policy based on the elapsed time since the last purchase or ad campaign. Moreover, we find that the seller tends to push ads earlier to customers with high purchase rates and low recapture rates (i.e., low transition rates from an inactive state to an active state). Surprisingly, we also find that the seller should push ads earlier to customers with intermediate churn rates (i.e., intermediate transition rates from an active state to an inactive state) compared to those with small or large churn rates. When confronted with a budget constraint on the overall ad cost, we suggest first solving a relaxed problem and then implementing a separate threshold policy for each customer with a specified budget. We establish the asymptotic optimality of this policy and show that the asymptotic loss order is O(1/\sqrt{T}). In addition, we explore the case with strategic customer reactions. Overall, our work reveals important insights of the optimal ad campaign problem and provides a useful strategy in practice.

链接：https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4477931

6. Regulating Discriminatory Pricing in the Presence of Tacit Collusion
作者：Zongsen Yang, Xiao Lei, Pin Gao

论文摘要：Price-setting algorithms have facilitated widespread awareness-based price discrimination, wherein firms charge high prices to customers unaware of alternative choices and low prices to those in competitive markets. This unethical behavior has increased customer complaints and prompted policymakers to enact fairness regulations in response. However, while limiting price discrimination may improve consumer welfare under genuine competition, it also affects firms' incentives to form tacit collusion, another regulatory concern arising from the proliferation of pricing algorithms. We develop an analytical model to examine the interplay between fairness regulation and tacit collusion, and discuss its impact on consumer welfare and policymaking. Firms utilize customers' product unawareness to implement price discrimination and decide whether to collude by comparing profits from collusion and deviation. We then explore the consequences of price fairness regulation on the sustainability of tacit collusion. For homogeneous products, fairness regulation can substantially weaken collusion, potentially rendering it unattainable. However, for differentiated products, strict fairness inadvertently supports collusive behavior, harming consumer welfare. In this case, mild fairness permitting moderate price differentiation can prevent market collusion and optimize welfare. Our key results are valid under various robustness checks such as incorporating Q-learning or using alternative demand models and punishment strategies. We also propose a novel randomized approach to tame fairness-induced collusion. Overall, our study emphasizes the importance of a nuanced approach to regulating discriminatory pricing in the presence of tacit collusion.

链接：https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4633784