BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Information theoretic\, message passing\, and spectral method for 
 inference in stochastic and censored block models
DTSTART:20160705T080000
DTEND:20160705T100000
DTSTAMP:20260408T034023Z
UID:9d681f4a93f37d4effe060ed847d0be142aa552af62681407e7a805b
CATEGORIES:Conferences - Seminars
DESCRIPTION:Chun Lam Chan\nEDIC Candidacy Exam\nExam President: Prof. Patr
 ick Thiran\nThesis Director: Dr. Nicolas Macris\nCo-examiner: Dr. Olivier 
 Leveque\nBackground papers:Asymptotic Mutual Information for the Two-Group
 s Stochastic Block Model\, by Y. Deshpande\; E. Abbe\; A. Montanari.Spectr
 al Detection in the Censored Block Model\, by Alaa Saade\; Florent Krzakal
 a\; Marc Lelarge\; Lenka Zdeborova.Spectral redemption: clustering sparse 
 networks\, by Florent Krzakala\; Cristopher Moore\; Elchanan Mossel\; Joe 
 Neeman\; Allan Sly\; Lenka\nZdeborova\; Pan Zhang.Abstract\nThe stochastic
  block model (SBM) and censored\nblock model (CBM) are two statistical mod
 els for the task of\nidentifying the hidden partition of vertices from an 
 undirected\ngraph. This problem is also known as community detection for\n
 network science. In this proposal\, we survey three papers that\nhave demo
 nstrated the use of theoretical tools listed in the title.\nIn a sufficien
 tly dense graph regime of SBM\, [1] has provided\nan explicit characteriza
 tion of the asymptotic mutual information\nand revealed a phase transition
  for partial recovery. In a sparse\ngraph regime\, [2] and [3] have propos
 ed spectral algorithms\nfor SBM and CBM\, respectively. The algorithms hav
 e an origin\nfrom message-passing algorithms and are observed to successfu
 lly\nprovide partial recovery as long as it is information-theoretically\n
 possible. Finally\, we discuss the connection of these inference\nproblems
  and coding theory.
LOCATION:BC 129 https://plan.epfl.ch/?room==BC%20129
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR