BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:IC Colloquium : Statistical and Computational Tradeoffs in High Di
 mensional Learning
DTSTART:20140313T161500
DTEND:20140313T173000
DTSTAMP:20260407T101157Z
UID:27e2815ee6198cd41bc7c8af2e6aa86e2d7fbe34766c45b1ee8f22bd
CATEGORIES:Conferences - Seminars
DESCRIPTION:By : Philippe Rigollet\, Princeton University\nIC Faculty cand
 idateAbstract\nComputational limitations of statistical problems have larg
 ely been ignored or simply overcome by ad hoc relaxations techniques. If o
 ptimal methods cannot be computed in reasonable time\, what is the best po
 ssible statistical performance of a computationally efficient procedure? B
 uilding on average case reductions\, we establish these fundamental limits
  in the context of sparse principal component analysis and quantify the st
 atistical price to pay for computational efficiency. Our results can be vi
 ewed as complexity theoretic lower bounds conditionally on the assumptions
  that some instances of the planted clique problem cannot be solved in ran
 domized polynomial time. [Joint work with Quentin Berthet]Biography\nPhili
 ppe Rigollet is an assistant professor in the department of Operations Res
 earch and Financial Engineering at Princeton University. His research inte
 rests focus on optimality of statistical methods and algorithms in the con
 text of high-dimensional statistical learning. His results draws connectio
 ns between optimization\, probability\, theoretical computer science and i
 nformation theory. His work is funded by several NSF grants including a CA
 REER award from the Division of Mathematical Sciences. He is also the reci
 pient of the Wentz award for outstanding research and teaching at Princeto
 n and of the Best Paper Award at the last Conference on Learning Theory (C
 OLT) for his work on statistical and computational tradeoffs.\nMore inform
 ation
LOCATION:BC 420 https://plan.epfl.ch/?room==BC%20420
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR