BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Learning determinantal point processes - Philippe Rigollet (Massac
 husetts Institute of Technology)
DTSTART:20180119T094500Z
DTEND:20180119T103000Z
UID:TALK97903@talks.cam.ac.uk
CONTACT:INI IT
DESCRIPTION:<span>Co-authors: Victor-Emmanuel Brunel		(MIT)\, Ankur Moitra
 		(MIT)\, John Urschel		(MIT)        <br></span><br>Determinantal Point Pr
 ocesses (DPPs) are a family of probabilistic models that have a repulsive 
 behavior\, and lend themselves naturally to many tasks in machine learning
  (such as recommendation systems) where returning a diverse set of objects
  is important. While there are fast algorithms for sampling\, marginalizat
 ion and conditioning\, much less is known about learning the parameters of
  a DPP. In this talk\, I will present recent results related to this probl
 em\, specifically - Rates of convergence for the maximum likelihood estima
 tor: by studying the local and global geometry of the expected log-likelih
 ood function we are able to establish rates of convergence for the MLE and
  give a complete characterization of the cases where these are parametric.
  We also give a partial description of the critical points for the expecte
 d log-likelihood. - Optimal rates of convergence for this problem: these a
 re achievable by the method of moments and are governed by a combinatorial
  parameter\, which we call the cycle sparsity. - A fast combinatorial algo
 rithm to implement the method of moments efficiently.  <br><br>The necessa
 ry background on DPPs will be given in the talk. <br><span><br>Joint work 
 with Victor-Emmanuel Brunel (M.I.T)\, Ankur Moitra (M.I.T) and John Ursche
 l (M.I.T).</span>
LOCATION:Seminar Room 1\, Newton Institute
END:VEVENT
END:VCALENDAR