BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Sentence Generation using a Dynamic Canvas - David Barber (Univers
 ity College London)
DTSTART:20171130T110000Z
DTEND:20171130T120000Z
UID:TALK96346@talks.cam.ac.uk
CONTACT:Dimitri Kartsaklis
DESCRIPTION:I will discuss the Attentive Unsupervised Text Writer (AUTR)\,
  a word level generative model for natural language. It uses a recurrent n
 eural network with a dynamic attention and canvas memory mechanism to iter
 atively construct sentences. By viewing the state of the memory at interme
 diate stages and where the model is placing its attention\, we gain insigh
 t into how it constructs sentences. We demonstrate that AUTR learns a usef
 ul latent representation for each sentence and achieves competitive log-li
 kelihood lower bounds whilst being computationally efficient. It is effect
 ive at generating and reconstructing sentences\, as well as imputing missi
 ng words.
LOCATION: SR-24\, English Faculty Building\, 9 West Road (Sidgwick Site)
END:VEVENT
END:VCALENDAR