BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Reduced plug-in rules for learning - Frederic Desobry (University 
 of Cambridge)
DTSTART:20080304T143000Z
DTEND:20080304T153000Z
UID:TALK9601@talks.cam.ac.uk
CONTACT:Minor Gordon
DESCRIPTION:Performance obtained via regularised risk minimisation-based a
 lgorithms have triggered their use as standard tools for many learning tas
 ks. However\, such algorithms can be found to be difficult to apply in spe
 cific situations\, e.g. when the experimental setup involves large dataset
 s. We study alternatives to this risk minimisation framework\, which are b
 ased on simple yet effective algorithmic designs.\nThe focus of this talk 
 will mainly be on statistical classification. The example of a methodology
  is derived\, to discriminate between two classes known through a set of d
 ata distributed according to different probability density functions. A de
 cision rule is built as the plug-in of a kernel rule\, defined on a small 
 subset of the learning set. This framework allows for fast yet accurate es
 timates of the optimal classification rule. Convergence sanity checks\, as
  well as initial simulation results\, will be presented.\nExtensions to ot
 her approaches will also be discussed.
LOCATION:Room FW11\, Computer Laboratory\, William Gates Building
END:VEVENT
END:VCALENDAR