BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Deep Learning in Practice - Nick Rogers\, Churchill College
DTSTART:20160120T190000Z
DTEND:20160120T194000Z
UID:TALK62540@talks.cam.ac.uk
CONTACT:Matthew Ireland
DESCRIPTION:This talk will explore some of the problems that exist in the 
 area of deep learning and some of the mathematical techniques that are use
 d to overcome these issues. We will begin by examining how the vanishing g
 radient problem threatens the very existence of deep networks and then loo
 k at how researchers working over the past decade have managed to overcome
  these issues by using a number of distinct methods to drastically improve
  the rate of learning and ultimately\, the resulting accuracy of these net
 works. We will investigate how using the cross-entropy cost function impro
 ves the learning speed and also show how using an alternative to sigmoid n
 eurons can avoid the problem of neuron saturation. We will also analyse th
 e motivation behind regularisation and show how it can be used to combat t
 he problem of overfitting. These techniques are used by pioneering neural 
 networks such as the winners of the Large Scale Visual Recognition Challen
 ge and can result in networks that achieve human levels of performance in 
 some tasks.
LOCATION:Wolfson Hall\, Churchill College
END:VEVENT
END:VCALENDAR