BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Rank-preserving optimization on the cone of  positive semidefinite
  matrices: a geometric approach.
DTSTART:20100427T101500
DTSTAMP:20260427T201440Z
UID:41ea9c63b52f559e9027b57cb4b2dd1cd1787f9a59f6c610ea216a29
CATEGORIES:Conferences - Seminars
DESCRIPTION:Prof. R. Sepulchre\, Systems and Modeling\, Dept of Elec. Eng.
  and Comp. Sci.\, University of Liege\, Belgium. \nThe talk is an introduc
 tion to a recent computational framework for optimization over the\nset of
  fixed rank positive semidefinite matrices. The foundation is geometric  a
 nd the motivation\nis algorithmic\, with a bias towards low-rank computati
 ons in large-scale problems.\nWe will describe two quotient riemannian geo
 metries that are rooted in\nclassical matrix factorizations and that  lead
  to rank-preserving efficient computations\nin the cone of symmetric posit
 ive definite matrices. The field of applications is vast\,\nand the talk w
 ill survey recent developments that illustrate the potential of the approa
 ch\nin large-scale computational problems encountered in control\, optimiz
 ation\, and machine learning.\nThe talk is introductory and requires no pa
 rticular background in riemannian geometry.
LOCATION:MEC2405
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR