BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Guessing Program Annotations with Probabilistic Inference - Aditya
  Nori\, Microsoft Researcher
DTSTART:20110406T140000Z
DTEND:20110406T150000Z
UID:TALK30453@talks.cam.ac.uk
CONTACT:Microsoft Research Cambridge Talks Admins
DESCRIPTION:Program analysis tools require annotations in order to be effe
 ctive. We present techniques to guess annotations using hints from the pro
 grammer. In our experience\, we find that it is easier for programmers to 
 give hints\, if we allow specifying facts that are likely to hold. For ins
 tance\, the programmer might state that if a pointer is probed inside a me
 thod and then accessed\, then it is very likely that the pointer has not b
 een validated on entry to the method. We propose using probabilistic infer
 ence to deal with such hints\, using probability as the uniform currency i
 n representing and manipulating uncertainty.  We give two examples from ou
 r experience:  (1) Merlin\, an inference engine for information flow annot
 ations\, and (2) Anek\, an inference engine for aliasing annotations known
  as access permissions. These experiences have led to a general framework 
 called BayeZ for guessing annotations via a combination of logical and pro
 babilistic inference. We present preliminary results for BayeZ together wi
 th applications beyond programming languages.
LOCATION:Small lecture theatre\, Microsoft Research Ltd\, 7 J J Thomson Av
 enue (Off Madingley Road)\, Cambridge
END:VEVENT
END:VCALENDAR