BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:IC Mondays seminars - Multi-task Learning: Theory and Practice
DTSTART:20100419T161500
DTSTAMP:20260407T175634Z
UID:824dd5c012ae81bb6b5245b4da0247468db66609f2a8a1697639b807
CATEGORIES:Conferences - Seminars
DESCRIPTION:Dr. Massimiliano Pontil\, University College London\, Computer
  Science Department \nAstract\nWe discuss the problem of estimating a stru
 ctured matrix with a large number of elements. A key motivation for this p
 roblem occurs in multi-task learning. In this case\, the columns of the ma
 trix correspond to the parameters of different regression or classificatio
 n tasks\, and there is structure due to relations between the tasks. We pr
 esent a general method to learn the tasks' parameters as well as their str
 ucture. Our approach is based on solving a convex optimization problem\, i
 nvolving a data term and a penalty term. We highlight different types of p
 enalty terms which are of practical and theoretical importance. They imple
 ment structural relations between the tasks and achieve a sparse represent
 ations of parameters. We address computational issues as well as the predi
 ctive performance of the method. Finally we discuss how these ideas can be
  extended to learn non-linear task functions by means of reproducing kerne
 ls.\n\nBiography\nMassimiliano Pontil is a Reader and EPSRC Advanced Resea
 rch Fellow in the Department of Computer Science at University College Lon
 don. His research interests are in the area of machine learning with a foc
 us on regularization methods\, convex optimization and statistical estimat
 ion. He received the equivalent of an MSc and a PhD in Physics from the Un
 iversity of Genova in 1994 and 1999\, respectively. He has also held visit
 ing positions at several universities and was a Postdoctoral Fellow at the
  Massachusetts Institute of Technology.
LOCATION:INM 202
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR