He Wang is an Assistant Professor and Colonel John B. Day Early Career Professor in the School of Industrial and Systems Engineering at Georgia Tech. His research interests include pricing and revenue management, supply chain, transportation, and machine learning. His works have received 1st place in INFORMS Junior Faculty Interest Group paper competition, Best Paper in Operation Research Award by the MSOM Society, NSF CAREER Award, and faculty research awards from Amazon and Didi.
| Venue | Seminar Room at Innovation 4.0 building (Level 1) |
| Link to Register
(Hybrid session) |
https://nus-sg.zoom.us/meeting/register/tZwudemurj4sHdy-N4Bl9b1CoBEnCO0vZ7mw |
| Title | Constant Regret Re-solving Heuristics for Revenue Management Problems |
| Abstract | We will discuss a classic network revenue management model of Gallego and van Ryzin (1997), which considers a retailer who sells a product (or multiple products) subject to initial inventory constraints over T consecutive periods. Because the optimal policy via dynamic programming is computationally intractable, researchers have proposed various approximate policies for this problem. We are interested in the so-called “re-solving heuristic,” which periodically solves the fluid approximation model. In the quantity-based revenue management setting with discrete types (joint work with P. Bumpensanti), we find that the re-solving heuristic has a worst-case regret of O(T^{1/2}) compared to the optimal policy, whereas a modified re-solving heuristic can achieve uniformly bounded O(1) regret. In the price-based revenue management setting with continuous price sets (joint work with Yining Wang), we show that the re-solving heuristic attains O(1) regret compared to the value of the optimal policy. This improves the O(lnT) regret upper bound established by Jasin (2014). In addition, we prove that there is an Ω(lnT) gap between the value of the optimal policy and that of the fluid model, implying that the fluid model is not an adequate benchmark for constant regret. |
