BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:"Transfer Learning for Scientific Discovery" by Dr. Alexandru Nicu
 lescu-Mizil\,  IBM T.J. Watson Research center
DTSTART:20100319T140000
DTSTAMP:20260408T144552Z
UID:6f178f63b008d3bb3b303e83e634c9e74427a322649603d0c7797c10
CATEGORIES:Conferences - Seminars
DESCRIPTION:Alexandru Niculescu-Mizil\, IBM T.J. Watson Research center\nA
 bstract\nWhile receiving significant attention\, learning Bayesian Network
  structure from data remains challenging\, especially when training data i
 s scarce. In this talk I show how structure learning performance can be si
 gnificantly improved through inductive transfer\, when data is available f
 or multiple related problems.  Departing from the traditional approach of 
 learning the structures for each problem in isolation\, I present a score 
 and search algorithm for jointly learning multiple Bayesian Networks that 
 improves the leaned structures by transferring useful information among th
 e related problems.\n\nBiography\nAlexandru Niculescu-Mizil is a Herman Go
 ldstine postdoctoral fellow at IBM T.J. Watson Research Center. He receive
 d his Ph.D. from Cornell University in 2008 under the supervision of Rich 
 Caruana\, a Masters of Science degree in Computer Science from Cornell Uni
 versity and a Magna Cum Laude Bachelors degree in Mathematics and Computer
  Science from University of Bucharest. His research interests are in machi
 ne learning and data mining\, particularly in inductive transfer\, graphic
 al model structure learning\, probability estimation\, empirical evaluatio
 ns\, ensemble methods and on-line learning. He received an ICML Distinguis
 hed Student Paper Award in 2005 for his work on probability estimation\, a
 nd a COLT Best Student in 2008 paper award for his work on on-line learnin
 g. In 2009 he led the IBM Research team that won the KDD Cup.
LOCATION:BC 01 https://plan.epfl.ch/?room==BC%2001
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR