BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:A clustering algorithm for Hüsler-Reiss graphical models
DTSTART:20260407T091500Z
DTEND:20260407T101500Z
DTSTAMP:20260410T114600Z
UID:indico-event-14476@indico.math.cnrs.fr
DESCRIPTION:Speakers: Marine Demangeot (Université Paul Valéry)\n\nExtre
 me events\, although rare\, can have substantial environmental\, societal\
 , and economic consequences. Their modeling is typically addressed within 
 the framework of extreme value theory\, where statistical inference is cha
 llenging due to the scarcity of observations\, especially in high dimensio
 ns. This motivates the search for parsimonious and interpretable structure
 s. Graphical models provide a natural framework by exploiting conditional 
 independence\, leading to sparse representations that facilitate interpret
 ation and estimation. In the Hüsler–Reiss setting\, this structure can 
 be captured through an extremal precision matrix\, whose zero pattern enco
 des the underlying graphical dependencies. In this work\, we propose a met
 hodology for Hüsler–Reiss graphical models that leverages structural sp
 arsity through clustering. Building on a clustering algorithm proposed by 
 Touw et al. [1]\, we introduce a convex fusion penalty on the extremal pre
 cision matrix during maximum likelihood estimation\, encouraging variables
  with similar tail dependence patterns to cluster together.  This approac
 h promotes block structures\, which in turn can induce conditional indepen
 dence. We study the theoretical properties of the method and prove consist
 ency of the estimator under a block-structured assumption. Simulation stud
 ies illustrate the performance of the method.\n[1] Touw\, D. J. W.\, Alfon
 s\, A.\, Groenen\, P. J. F.\, & Wilms\, I. (2026). Clusterpath Gaussian Gr
 aphical Modeling. arXiv preprint arXiv:2407.00644\nJoint work with Alexand
 re Capel\, Nicolas Meyer and Gwladys Toulemonde (Université de Montpellie
 r\, IMAG)\n\nhttps://indico.math.cnrs.fr/event/14476/
LOCATION:Salle K. Johnson (1R3\, 1er étage)
URL:https://indico.math.cnrs.fr/event/14476/
END:VEVENT
END:VCALENDAR