BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//132.216.98.100//NONSGML kigkonsult.se iCalcreator 2.20.4//
BEGIN:VEVENT
UID:20260601T131220EDT-7578jrXpjL@132.216.98.100
DTSTAMP:20260601T171220Z
DESCRIPTION:Title: Feature Selection for Linear Fixed Effects Models\n\n \n
 \nAbstract: \n\nLinear mixed-effects (LME) models are used to analyze nest
 ed or combined data across a range of groups or clusters. These models use
  covariates to separate the total population variability (the fixed effect
 s) from the group variability (the random effects). LMEs borrow strength a
 cross groups to estimate key statistics in cases where the data within gro
 ups may be sparse or highly variable\, and play a fundamental role in popu
 lation health sciences\, meta-analysis\, life sciences\, and in many other
 s domains. In this talk we formally introduce a mathematical description o
 f the LME model and its feature selection variant. A naive proximal gradie
 nt descent (PGD) algorithm for its solution is described and its deficienc
 ies are explained. A novel solution strategy is proposed that is based on 
 relaxation strategy that decouples the smooth from the nonsmooth component
 s of the maximum likelihood objective. An optimal value function is obtain
 ed by partially optimizing the smooth component of the decoupled problem. 
 We show that the resulting optimal value function has a locally Lipschitz 
 gradient and so a PGD algorithm can be applied to a feature selecting regu
 larization of the optimal value function. At first this approach seems cou
 nter intuitive since the optimal value function adds yet another layer of 
 complexity to the problem. However\, this complexity is mitigated by the u
 se of modern variational and numerical techniques. The resulting PGD algor
 ithm applied to this reformulation is more stable and can rapidly identify
  the important features to high accuracy. The algorithmic details and the 
 numerical results are presented.\n\nLocation\n\nCentre de recherches mathé
 matiques Pavillon André-Aisenstadt\, Université de Montréal Room 5340\n
DTSTART:20231104T200000Z
DTEND:20231104T210000Z
SUMMARY:James V. Burke (University of Washington)
URL:/mathstat/channels/event/james-v-burke-university-
 washington-352344
END:VEVENT
END:VCALENDAR