BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:AI+Pizza January 2018 - Microsft Research Cambridge/University of 
 Cambridge
DTSTART:20180119T173000Z
DTEND:20180119T190000Z
UID:TALK97864@talks.cam.ac.uk
CONTACT:Microsoft Research Cambridge Talks Admins
DESCRIPTION:*Speaker 1:* \nMiltos Allamanis (MSR Cambridge).\n*Title:* \nL
 earning to Detect Bugs in Source Code with Machine Learning.\n*Abstract:*\
 nDeep neural networks are succeeding at a range of natural language tasks 
 such as machine translation and text summarization. Recently\, the interdi
 sciplinary field of "big code" promises a new set of learnable statistical
  static analyses. While machine learning tasks on source code have been co
 nsidered recently\, most work in this area does not attempt to capitalize 
 on the unique opportunities offered by its known syntax and structure. In 
 this talk\, I discuss how graph neural networks can learn from code's synt
 actic and semantic structure to detect variable misuse bugs in code withou
 t any external information (e.g. unit tests).\n(Joint work with Marc Brock
 schmidt and others).\n\n*Speaker 2:*\nNiki Kilbertus (Cambridge university
 ).\n*Title:*\nLearning Independent Causal Mechanisms.\n*Abstract:* \nIndep
 endent causal mechanisms are a central concept in the study of causality w
 ith implications for machine learning tasks. In this work we develop an al
 gorithm to recover a set of (inverse) independent mechanisms relating a di
 stribution transformed by the mechanisms to a reference distribution. The 
 approach is fully unsupervised and based on a set of experts that compete 
 for data to specialize and extract the mechanisms. We test and analyze the
  proposed method on a series of experiments based on image transformations
 . Each expert successfully maps a subset of the transformed data to the or
 iginal domain\, and the learned mechanisms generalize to other domains. We
  discuss implications for domain transfer and links to recent trends in ge
 nerative modeling.\n(Joint work with Giambattista Parascandolo\, Mateo Roj
 as-Carulla\, and Bernhard Schölkopf).\n
LOCATION:Auditorium\, Microsoft Research Ltd\, 21 Station Road\, Cambridge
 \, CB1 2FB
END:VEVENT
END:VCALENDAR