BEGIN:VCALENDAR
VERSION:2.0
PRODID:IEEE vTools.Events//EN
CALSCALE:GREGORIAN
BEGIN:VTIMEZONE
TZID:America/Chicago
BEGIN:DAYLIGHT
DTSTART:20240310T030000
TZOFFSETFROM:-0600
TZOFFSETTO:-0500
RRULE:FREQ=YEARLY;BYDAY=2SU;BYMONTH=3
TZNAME:CDT
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:20231105T010000
TZOFFSETFROM:-0500
TZOFFSETTO:-0600
RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=11
TZNAME:CST
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20231115T021824Z
UID:354AE90B-8C4F-4165-B146-80407E4922D1
DTSTART;TZID=America/Chicago:20231114T110000
DTEND;TZID=America/Chicago:20231114T123000
DESCRIPTION:Title:  Statistical Inference over Networks: Decentralized Opti
 mization Meets High-dimensional Statistics\n\nThere is growing interest in
  solving large-scale statistical machine learning problems over decentrali
 zed networks\, where 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 cha
 llenge at the nexus of the computational and statistical sciences: ensurin
 g the quality of statistical inference when computational resources\, like
  time and communication\, are constrained. While statistical-computation t
 radeoffs have been largely explored in the centralized setting\, our under
 standing over mesh networks is limited: (i) distributed schemes\, designed
  and performing well in the classical low-dimensional regime\, can break d
 own in the high-dimensional case\; and (ii) existing convergence studies m
 ay fail to predict algorithmic behaviors\, with some findings directly con
 tradicted by empirical tests. This is mainly due to the fact that the majo
 rity of distributed algorithms have been designed and studied only from th
 e optimization perspective\, lacking the statistical dimension. This talk 
 will discuss some vignettes from high-dimensional statistical inference su
 ggesting new analyses (and designs) aiming at bringing statistical thinkin
 g in distributed optimization.\n\nCo-sponsored by: College of Science and 
 Engineering\, University of Minnesota\n\nSpeaker(s): Prof. Gesualdo Scutar
 i\n\nAgenda: \n11:00 – 11:05 p.m. Welcome Remarks by Dr. Wenyu Jin\n\n11
 :05 – 12:00 p.m.   Statistical Inference over Networks: Decentralized Op
 timization Meets High-dimensional Statistics by Prof. Gesualdo Scutari \, 
 chair by Prof. Mingyi Hong\n\n12:00 – 12:30 p.m. Q&A and Networking\, Pr
 of. Mingyi Hong to Moderate\n\nRoom: 3-180\, Bldg: Kenneth H. Keller Hall\
 , 200 Union St SE\, Minneapolis\, Minnesota\, United States\, 55455
LOCATION:Room: 3-180\, Bldg: Kenneth H. Keller Hall\, 200 Union St SE\, Min
 neapolis\, Minnesota\, United States\, 55455
ORGANIZER:wenyu.jin@ieee.org
SEQUENCE:47
SUMMARY:IEEE SPS Distinguished Lecturer Program Twin Cities SP/COM Chapter 
 Seminar 11/14/2023
URL;VALUE=URI:https://events.vtools.ieee.org/m/376976
X-ALT-DESC:Description: <br /><p>Title: <strong><span style="text-decoratio
 n: underline\;">&nbsp\;Statistical Inference over Networks: Decentralized 
 Optimization Meets High-dimensional&nbsp\;&nbsp\;Statistics</span></strong
 ></p>\n<p>There is growing interest in solving large-scale statistical mac
 hine learning problems over decentralized networks\, where data are distri
 buted across the nodes of the network and no centralized coordination is p
 resent (we termed these systems as &ldquo\;mesh&rdquo\; networks). Inferen
 ce from massive datasets poses&nbsp\;&nbsp\;a fundamental challenge at the
  nexus of the computational and statistical sciences: ensuring the quality
  of statistical inference when computational resources\, like time and com
 munication\, are constrained.&nbsp\; &nbsp\;While statistical-computation 
 tradeoffs have been largely explored in the centralized setting\, our unde
 rstanding over mesh networks is limited: (i) distributed schemes\, designe
 d and performing well in the classical low-dimensional regime\, can break 
 down in the high-dimensional case\; and (ii) existing convergence studies 
 may fail to predict algorithmic behaviors\, with some findings directly co
 ntradicted by empirical tests. This is mainly due to the fact that the maj
 ority of distributed algorithms&nbsp\;&nbsp\;have been designed and studie
 d only from the optimization perspective\, lacking the statistical dimensi
 on. This talk will discuss some vignettes from&nbsp\;&nbsp\;high-dimension
 al statistical inference suggesting&nbsp\;&nbsp\;new analyses (and designs
 ) aiming at bringing statistical thinking in distributed optimization.</p>
 <br /><br />Agenda: <br /><p>11:00 &ndash\; 11:05 p.m.&nbsp\;&nbsp\;&nbsp\
 ;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\; <strong>Welcome Remarks</stron
 g>&nbsp\;by&nbsp\;<em>Dr. Wenyu Jin</em></p>\n<p>11:05 &ndash\; 12:00 p.m.
  &nbsp\; &nbsp\; &nbsp\; &nbsp\;<strong> <span style="text-decoration: und
 erline\;">&nbsp\;Statistical Inference over Networks: Decentralized Optimi
 zation Meets High-dimensional&nbsp\;&nbsp\;Statistics</span></strong> by P
 rof.&nbsp\; Gesualdo Scutari \, chair by Prof. Mingyi Hong</p>\n<p>12:00 &
 ndash\; 12:30 p.m.&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;&nbsp\;
 &nbsp\; <strong>Q&amp\;A and Networking\, </strong><em>Prof. Mingyi Hong t
 o Moderate</em></p>\n<p>&nbsp\;</p>
END:VEVENT
END:VCALENDAR