BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//132.216.98.100//NONSGML kigkonsult.se iCalcreator 2.20.4//
BEGIN:VEVENT
UID:20250511T123050EDT-9747rUnMXb@132.216.98.100
DTSTAMP:20250511T163050Z
DESCRIPTION:Ming Hu\n\nUniversity of Toronto\n	Rotman School of Management\n
 \nBig\, Small\, and Small+ Data-Driven OM\n\nDate: Friday\, November 15\, 
 2024\n	Time: 10:00 - 11:00 am\n	Location: Bronfman building\, Room 045\n\n\n
 Abstract\n\nWe discuss three data-driven decision-making scenarios in clas
 sical operations management (inventory or pricing) settings\, using big or
  small data or small data but with one chance of experimentation. First\, 
 with big data\, we study a contextual-based newsvendor problem using deep 
 neural networks (DNN). Empirical process theory is pivotal in ensuring tha
 t the asymptotic behavior of observed data converges to the true underlyin
 g distribution as the sample size increases. We provide theoretical guaran
 tees in terms of excess risk bounds for the DNN solution\, characterized b
 y the network structure and sample size\, validating the applicability of 
 DNNs in relevant OM contexts. These excess risk bounds exhibit polynomial 
 growth in the feature dimension and attain the minimax convergence rate (w
 ith respect to the sample size) in expectation. Second\, with small data\,
  traditional frequentist methods may be ineffective\, and we propose using
  the empirical Bayes (EB) method to achieve transfer learning in estimatin
 g unknown parameters using data across many products and subsequently maki
 ng decisions based on these estimates. We illustrate this approach with a 
 multi-product pricing problem\, employing a hierarchical feature-based dem
 and model and a nonparametric maximum likelihood method to derive the prio
 r from the data. The effectiveness of the EB method is demonstrated by cha
 racterizing the regret bound using an oracle benchmark that presumes prior
  knowledge of the underlying distribution. Third\, we study the benefits o
 f a one-shot price experiment for a seller in setting a price who only kno
 ws the exact purchase probability associated with a single historical pric
 e and aims to maximize the worst-case revenue ratio compared to an oracle 
 with complete knowledge of the value distribution. We analytically charact
 erize the optimal distributionally robust experimental and final price poi
 nts\, obtain their tight performance guarantee for any historical purchase
  probability\, and then evaluate the value of a one-shot experiment\, whic
 h exhibits a bimodal behavior with respect to the historical purchase prob
 ability.\n\nBio\n\nMing Hu is the University of Toronto Distinguished Prof
 essor of Business Operations and Analytics\, a professor of operations man
 agement at the Rotman School of Management\, and an Amazon Scholar. He rec
 eived a master's degree in Applied Mathematics from Brown University in 20
 03 and a Ph.D. in Operations Research from Columbia University in 2009.\n
DTSTART:20241115T150000Z
DTEND:20241115T160000Z
LOCATION:Room 045\, Bronfman Building\, CA\, QC\, Montreal\, H3A 1G5\, 1001
  rue Sherbrooke Ouest
SUMMARY:Management Science Research Centre (MSRC) Seminar: Ming Hu
URL:/desautels/channels/event/management-science-resea
 rch-centre-msrc-seminar-ming-hu-360642
END:VEVENT
END:VCALENDAR