BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Variational Smoothing in Recurrent Neural Network Language Models 
 - Dr. Lingpeng Kong (DeepMind)
DTSTART:20200227T110000Z
DTEND:20200227T120000Z
UID:TALK140221@talks.cam.ac.uk
CONTACT:Qianchu Liu
DESCRIPTION:In this talk\, we present a new theoretical perspective of dat
 a noising in recurrent neural network language models (Xie et al.\, 2017).
  We show that each variant of data noising is an instance of Bayesian recu
 rrent neural networks with a particular variational distribution (i.e.\, a
  mixture of Gaussians whose weights depend on statistics derived from the 
 corpus such as the unigram distribution). We use this insight to propose a
  more principled method to apply at prediction time and propose natural ex
 tensions to data noising under the variational framework. In particular\, 
 we propose variational smoothing with tied input and output embedding matr
 ices and an element-wise variational smoothing method. We empirically veri
 fy our analysis on two bench-mark language modeling datasets and demonstra
 te performance improvements over existing data noising methods.
LOCATION:Board room\, Faculty of English\, 9 West Rd (Sidgwick Site)
END:VEVENT
END:VCALENDAR