BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//132.216.98.100//NONSGML kigkonsult.se iCalcreator 2.20.4//
BEGIN:VEVENT
UID:20260602T200405EDT-1861BKJufH@132.216.98.100
DTSTAMP:20260603T000405Z
DESCRIPTION:Title: Learning from a Biased Sample\n\nAbstract: The empirical
  risk minimization approach to data-driven decision making assumes that we
  can learn a decision rule from training data drawn under the same conditi
 ons as the ones we want to deploy it under. However\, in a number of setti
 ngs\, we may be concerned that our training sample is biased\, and that so
 me groups (characterized by either observable or unobservable attributes) 
 may be under- or over-represented relative to the general population\; and
  in this setting empirical risk minimization over the training set may fai
 l to yield rules that perform well at deployment. Building on concepts fro
 m distributionally robust optimization and sensitivity analysis\, we propo
 se a method for learning a decision rule that minimizes the worst-case ris
 k incurred under a family of test distributions whose conditional distribu
 tions of outcomes  given covariates  differ from the conditional training 
 distribution by at most a constant factor\, and whose covariate distributi
 ons are absolutely continuous with respect to the covariate distribution o
 f the training data. We apply a result of Rockafellar and Uryasev to show 
 that this problem is equivalent to an augmented convex risk minimization p
 roblem. We give statistical guarantees for learning a robust model using t
 he method of sieves and propose a deep learning algorithm whose loss funct
 ion captures our robustness target. We empirically validate our proposed m
 ethod in simulations and a case study with the MIMIC-III dataset.\n	\n	Paper
  link: https://arxiv.org/abs/2209.01754\n\nWeb site : https://mcgillstat.g
 ithub.io\n
DTSTART:20230203T203000Z
DTEND:20230203T213000Z
SUMMARY:Lihua Lei\, Stanford Graduate School of Business
URL:/mathstat/channels/event/lihua-lei-stanford-gradua
 te-school-business-345745
END:VEVENT
END:VCALENDAR