BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Boolean interactions: practice and theory
DTSTART:20230428T151500
DTEND:20230428T170000
DTSTAMP:20260406T230206Z
UID:774902daae96f8b0959d75af35f434876e1be17cf05d26d193273536
CATEGORIES:Conferences - Seminars
DESCRIPTION:Bin Yu\, UC Berkeley Statistics\nThresholding or Boolean behav
 iors of biomolecules underlie many biological processes.\nDecision-trees c
 apture such behaviors and tree-based methods such random forests have been
  shown to succeed in predictive tasks in genomics and medicine.\nIn this t
 alk\, we use UKBiobank data and a stable version of the random forests\, i
 terative random forests (iRF)\, to recommend gene and gene-gene interactio
 ns that have data evidence for driving a heart disease\, Hypertrophic Card
 iomyopathy (HCM).\n4 out of the 5 recommendations are shown to be causal f
 or HCM in follow-up gene-silencing experiments. This and other empirical s
 uccesses of iRF motivate a theoretical investigation of its tractable vers
 ion under a new local sparse and spiky (LSS) model where the regression fu
 nction is a linear combination of Boolean interactions of features.\nThe t
 ractable version of iRF is shown to be model selection consistent under th
 is new model with conditions of feature independence and non-overlapping i
 nteractions.\n 
LOCATION:GA 3 21 https://plan.epfl.ch/?room==GA%203%2021 https://epfl.zoom
 .us/j/63996788598
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR