BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Differentially Private Bayesian Learning - Dr Antti Honkela\, Univ
 ersity of Helsinki
DTSTART:20170314T110000Z
DTEND:20170314T120000Z
UID:TALK71519@talks.cam.ac.uk
CONTACT:Alexander Matthews
DESCRIPTION:Many applications of machine learning for example in health ca
 re would benefit from methods that can guarantee data subject privacy.Diff
 erential privacy has recently emerged as a leading framework for private d
 ata analysis. Differential privacy guarantees privacy by requiring that th
 e results of an algorithm should not change much even if one data point is
  changed\, thus providing plausible deniability for the data subjects.\n\n
 In this talk I will present methods for efficient differentially private B
 ayesian learning. In addition to asymptotic efficiency\, we will focus on 
 how to make the methods efficient for moderately-sized data sets. The meth
 ods are based on perturbation of sufficient statistics for exponential fam
 ily models and perturbation of gradients for variational inference. Unlike
  previous state-of-the-art\, our methods can predict drug sensitivity of c
 ancer cell lines using differentially private linear regression with bette
 r accuracy than using a very small non-private data set.
LOCATION:CBL Room BE-438\, Department of Engineering
END:VEVENT
END:VCALENDAR