BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Journal Club: &quot\;Reducing the Dimensionality of Data with Neur
 al Networks&quot\; - O. Stegle
DTSTART:20061208T123000Z
DTEND:20061208T133000Z
UID:TALK6032@talks.cam.ac.uk
CONTACT:Oliver Stegle
DESCRIPTION:G. E. Hinton* and R. R. Salakhutdinov\nHigh-dimensional data c
 an be converted to low-dimensional codes by training a multilayer\nnetwork
  with a small central layer to reconstruct high-dimensional input vectors.
  Gradient descent\ncan be used for fine-tuning the weights in such ‘‘a
 utoencoder’’ networks\, but this works well\nthe initial weights are c
 lose to a good solution. We describe an effective way of initializing\nwei
 ghts that allows deep autoencoder networks to learn low-dimensional codes 
 that work much\nbetter than principal components analysis as a tool to red
 uce the dimensionality of data.\n----\nhttp://www.sciencemag.org/cgi/repri
 nt/313/5786/504.pdf\n\nplease also look at supporting online Material:\n\n
 http://www.sciencemag.org/cgi/data/313/5786/504/DC1/1\n\n
LOCATION:TCM Seminar Room\, Cavendish Laboratory\, Department of Physics
END:VEVENT
END:VCALENDAR