喻园管理论坛2023年第32期(总第855期)

演讲主题: Efficient Global Converging Algorithms for Booking Limit Control in Nonconvex Network Revenue Management

主 讲 人: 胡逸凡，瑞士洛桑联邦理工学院博士后

主 持 人: 李   锋，管理学院生产运作与物流管理系副教授

活动时间: 2023年4月25日（周二）10:00-11:30

活动地点: 管理大楼402教室

主讲人简介：

Yifan Hu is a postdoctoral researcher in the Risk Analytics and Optimization Lab at EPFL, jointly advised by Prof. Daniel Kuhn and Prof. Andreas Krause. Prior to that, he obtained Doctor in Industrial Engineering from UIUC, jointly advised by Prof. Xin Chen and Prof. Niao He. His research interest lies in building theoretical understanding of optimization, operations research, and machine learning. Specifically, he is interested in designing efficient algorithms for stochastic optimization, robust learning, bandit and reinforcement learning as well as solving operations problems i.e., revenue and inventory management with provable guarantees.

活动简介：

In this talk, we formulate booking limits control for air-cargo network revenue management (NRM) problem with random two-dimensional capacity, random consumption, and routing flexibility as a two-stage stochastic programming. Such problems belong to a general class of stochastic nonconvex optimization in the form min_x E[f(x∧ξ)], where in NRM f is convex, ξ is the random demand from an unknown distribution that truncates the booking limit decision x. To address the nonconvexity due to the truncation, we leverage an implicit convex reformulation via a variable transformation u = E[x∧ξ] and develop global converging stochastic gradient-based algorithms. Interestingly, our proposed Mirror Stochastic Gradient (MSG) method operates only in the original x-space using gradient estimators of the original nonconvex objective and achieves near-optimal sample and gradient complexities. Extensive numerical experiments demonstrate the superior performance of our proposed MSG algorithm for booking limit control with higher revenue and lower computation cost than state-of-the-art bid-price-based control policies, especially when the variance of random capacity is large.