BEGIN:VCALENDAR
VERSION:2.0
PRODID:IEEE vTools.Events//EN
CALSCALE:GREGORIAN
BEGIN:VTIMEZONE
TZID:US/Eastern
BEGIN:DAYLIGHT
DTSTART:20240310T030000
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
RRULE:FREQ=YEARLY;BYDAY=2SU;BYMONTH=3
TZNAME:EDT
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:20231105T010000
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=11
TZNAME:EST
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20231203T155423Z
UID:705F1486-5FD0-4E2D-B7EA-5FF36CD13E3C
DTSTART;TZID=US/Eastern:20231129T120000
DTEND;TZID=US/Eastern:20231129T130000
DESCRIPTION:There is growing interest in solving large-scale statistical ma
 chine learning problems over decentralized networks\, where data are distr
 ibuted across the nodes of the network and no centralized coordination is 
 present (we termed these systems as “mesh” networks). Inference from m
 assive datasets poses a fundamental challenge at the nexus of the computat
 ional and statistical sciences: ensuring the quality of statistical infere
 nce when computational resources\, like time and communication\, are const
 rained. While statistical-computation tradeoffs have been largely explored
  in the centralized setting\, our understanding over mesh networks is limi
 ted: (i) distributed schemes\, designed and performing well in the classic
 al low-dimensional regime\, can break down in the high-dimensional case\; 
 and (ii) existing convergence studies may fail to predict algorithmic beha
 viors\, with some findings directly contradicted by empirical tests. This 
 is mainly due to the fact that the majority of distributed algorithms have
  been designed and studied only from the optimization perspective\, lackin
 g the statistical dimension. This talk will discuss some vignettes from hi
 gh-dimensional statistical inference suggesting new analyses (and designs)
  aiming at bringing statistical thinking in distributed optimization.\n\nC
 o-sponsored by: Fairleigh Dickinson University\n\nSpeaker(s): Dr. Gesualdo
  Scutari  \, \n\nAgenda: \nThere is growing interest in solving large-scal
 e statistical machine learning problems over decentralized networks\, wher
 e data are distributed across the nodes of the network and no centralized 
 coordination is present (we termed these systems as “mesh” networks). 
 Inference from massive datasets poses a fundamental challenge at the nexus
  of the computational and statistical sciences: ensuring the quality of st
 atistical inference when computational resources\, like time and communica
 tion\, are constrained. While statistical-computation tradeoffs have been 
 largely explored in the centralized setting\, our understanding over mesh 
 networks is limited: (i) distributed schemes\, designed and performing wel
 l in the classical low-dimensional regime\, can break down in the high-dim
 ensional case\; and (ii) existing convergence studies may fail to predict 
 algorithmic behaviors\, with some findings directly contradicted by empiri
 cal tests. This is mainly due to the fact that the majority of distributed
  algorithms have been designed and studied only from the optimization pers
 pective\, lacking the statistical dimension. This talk will discuss some v
 ignettes from high-dimensional statistical inference suggesting new analys
 es (and designs) aiming at bringing statistical thinking in distributed op
 timization.\n\nVirtual: https://events.vtools.ieee.org/m/380601
LOCATION:Virtual: https://events.vtools.ieee.org/m/380601
ORGANIZER:tan@fdu.edu
SEQUENCE:23
SUMMARY:Statistical Inference over Networks: Decentralized Optimization Mee
 ts High-dimensional Statistics
URL;VALUE=URI:https://events.vtools.ieee.org/m/380601
X-ALT-DESC:Description: <br /><p>There is growing interest in solving large
 -scale statistical machine learning problems over decentralized networks\,
  where data are distributed across the nodes of the network and no central
 ized coordination is present (we termed these systems as &ldquo\;mesh&rdqu
 o\; networks). Inference from massive datasets poses&nbsp\;&nbsp\;a fundam
 ental challenge at the nexus of the computational and statistical sciences
 : ensuring the quality of statistical inference when computational resourc
 es\, like time and communication\, are constrained.&nbsp\; &nbsp\;While st
 atistical-computation tradeoffs have been largely explored in the centrali
 zed setting\, our understanding over mesh networks is limited: (i) distrib
 uted schemes\, designed and performing well in the classical low-dimension
 al regime\, can break down in the high-dimensional case\; and (ii) existin
 g convergence studies may fail to predict algorithmic behaviors\, with som
 e findings directly contradicted by empirical tests. This is mainly due to
  the fact that the majority of distributed algorithms&nbsp\;&nbsp\;have be
 en designed and studied only from the optimization perspective\, lacking t
 he statistical dimension. This talk will discuss some vignettes from&nbsp\
 ;&nbsp\;high-dimensional statistical inference suggesting&nbsp\;&nbsp\;new
  analyses (and designs) aiming at bringing statistical thinking in distrib
 uted optimization.<br /><br /></p><br /><br />Agenda: <br /><p>There is gr
 owing interest in solving large-scale statistical machine learning problem
 s over decentralized networks\, where data are distributed across the node
 s of the network and no centralized coordination is present (we termed the
 se systems as &ldquo\;mesh&rdquo\; networks). Inference from massive datas
 ets poses&nbsp\;&nbsp\;a fundamental challenge at the nexus of the computa
 tional and statistical sciences: ensuring the quality of statistical infer
 ence when computational resources\, like time and communication\, are cons
 trained.&nbsp\; &nbsp\;While statistical-computation tradeoffs have been l
 argely explored in the centralized setting\, our understanding over mesh n
 etworks is limited: (i) distributed schemes\, designed and performing well
  in the classical low-dimensional regime\, can break down in the high-dime
 nsional case\; and (ii) existing convergence studies may fail to predict a
 lgorithmic behaviors\, with some findings directly contradicted by empiric
 al tests. This is mainly due to the fact that the majority of distributed 
 algorithms&nbsp\;&nbsp\;have been designed and studied only from the optim
 ization perspective\, lacking the statistical dimension. This talk will di
 scuss some vignettes from&nbsp\;&nbsp\;high-dimensional statistical infere
 nce suggesting&nbsp\;&nbsp\;new analyses (and designs) aiming at bringing 
 statistical thinking in distributed optimization.<br /><br /></p>
END:VEVENT
END:VCALENDAR