BEGIN:VCALENDAR
VERSION:2.0
CALSCALE:GREGORIAN
METHOD:PUBLISH
X-WR-CALNAME:
X-WR-TIMEZONE:Europe/Paris
BEGIN:VTIMEZONE
TZID:Europe/Paris
X-LIC-LOCATION:Europe/Paris
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:19700329T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:19701025T030000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
SUMMARY:
A domain theory for statistical probabilistic programming
STATUS:CONFIRMED
ATTENDEE;CN="Ohad Kammar
":
MAILTO:no@spam.com
DESCRIPTION:
I will describe our recent work on statistical probabilisti
c programming languages. These are expressive languages for
describing generative Bayesian models of the kinds used in c
omputational statistics and machine learning. We give an ade
quate denotational semantics for a calculus with recursive h
igher-order types\, continuous probability distributions\, a
nd soft constraints. Among them are untyped languages\, simi
lar to Church and WebPPL\, because our semantics allows recu
rsive mixed-variance datatypes. Our semantics justifies impo
rtant program equivalences including commutativity. Our new
semantic model is based on `quasi-Borel predomains'. These a
re a mixture of chain-complete partial orders (cpos) and qua
si-Borel spaces. Quasi-Borel spaces are a recent model of pr
obability theory that focuses on sets of admissible random e
lements. I will give a brief introduction to quasi-Borel spa
ces and predomains\, and their properties. Probability is tr
aditionally treated in cpo models using probabilistic powerd
omains\, but these are not known to be commutative on any cl
ass of cpos with higher-order functions. By contrast\, quasi
-Borel predomains do support both a commutative probabilisti
c powerdomain and higher-order functions\, which I will desc
ribe. For more details on this joint work with Matthijs VÃ¡k
Ã¡r and Sam Staton\, see: Matthijs VÃ¡kÃ¡r\, Ohad Kammar\, a
nd Sam Staton. 2019. A Domain Theory for Statistical Probabi
listic Programming. Proc. ACM Program. Lang. 3\, POPL\, Arti
cle 36 (January 2019)\, 35 pages.\, DOI: 10.1145/3290349.
DTSTART;TZID=Europe/Paris:20190129T110000
DURATION:PT1H
URL;VALUE=URI:http://www.lsv.ens-cachan.fr/Seminaires/?sem=201901291
100
UID:LSVsemLSV.201901291100@lsv.ens-cachan.fr
LOCATION:Pavillon des Jardins
END:VEVENT
END:VCALENDAR