BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Inferring graph structure from signal observations
DTSTART:20160825T110000
DTEND:20160825T130000
DTSTAMP:20260407T182235Z
UID:8ac35547611dc323de865e3389468225333641888405f16064c6592f
CATEGORIES:Conferences - Seminars
DESCRIPTION:Hermina Petric Maretic\nEDIC Candidacy Exam\nExam President: P
 rof. Pierre Vandergheynst\nThesis Director: Prof. Pascal Frossard\nCo-exam
 iner: Prof. Daniel Kressner\nBackground papers:Learning Laplacian Matrix i
 n Smooth Graph Signal Representations\, by X. Dong\, et al.Fitting a Graph
  to Vector Data\, by S.Daitch\, et al.Efficient Dimensionality Reduction f
 or High-Dimensional Network Estimation\, by S. Celik\, et al.\nAbstract\nN
 etwork-structured data appears naturally in a large and constantly increas
 ing number of domains\, which makes its analysis crucial. However\, in ord
 er to successfully process data on graphs\, it is essential to place them 
 on a meaningful graph structure. As these structures are often not known o
 r uniquely defined\, we encounter the problem of inferring relevant graphs
  from signals.\nIn this proposal\, we discuss three different approaches t
 o the problem of graph learning. We first examine a more traditional appro
 ach from the machine learning community\, modelling network structure to g
 et further insight in solving machine learning problems. We then take a lo
 ok at a signal processing approach\, based on an extension of the classica
 l factor analysis model\, leveraging the connection of the graph Fourier t
 ransform to its topology. Finally\, we consider an approach relying on a p
 robabilistic model to learn large and highly structured graphs. We finish 
 with a short overview of our current work and a discussion of potential fu
 rther directions.
LOCATION:BC 410 https://plan.epfl.ch/?room==BC%20410
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR