BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//talks.cam.ac.uk//v3//EN
BEGIN:VTIMEZONE
TZID:Europe/London
BEGIN:DAYLIGHT
TZOFFSETFROM:+0000
TZOFFSETTO:+0100
TZNAME:BST
DTSTART:19700329T010000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0100
TZOFFSETTO:+0000
TZNAME:GMT
DTSTART:19701025T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
CATEGORIES:Machine Intelligence Lab Seminar
SUMMARY:Some new approaches to speech recognition and lang
 uage modelling - Geoffrey Hinton (University of To
 ronto)
DTSTART;TZID=Europe/London:20110622T140000
DTEND;TZID=Europe/London:20110622T150000
UID:TALK31797AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/31797
DESCRIPTION:In the first part of the talk\, I will review some
  of the research at the University of Toronto on t
 he use of deep belief nets for speech recognition.
  Then I will describe some very recent work that u
 ses a rather different approach to learn acoustic 
 events that have precise times and amplitudes. Fin
 ally I will briefly describe a much better way of 
 optimizing recurrent neural networks. When applied
  to a novel type of recurrent neural net\, this ne
 w optimizer learns extremely good models of charac
 ter sequences. This is joint work with George Dahl
 \, Abdel-rahman Mohamed\, Navdeep Jaitly\, James M
 artens and Ilya Sutskever.
LOCATION:Cambridge University Engineering Department\, Lect
 ure Room 2
CONTACT:Kai Yu
END:VEVENT
END:VCALENDAR