BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//talks.cam.ac.uk//v3//EN
BEGIN:VTIMEZONE
TZID:Europe/London
BEGIN:DAYLIGHT
TZOFFSETFROM:+0000
TZOFFSETTO:+0100
TZNAME:BST
DTSTART:19700329T010000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0100
TZOFFSETTO:+0000
TZNAME:GMT
DTSTART:19701025T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
CATEGORIES:Statistics
SUMMARY:It is hard to be strongly faithful - Caroline Uhle
r\, Institute of Science and Technology Austria
DTSTART;TZID=Europe/London:20140502T160000
DTEND;TZID=Europe/London:20140502T170000
UID:TALK52179AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/52179
DESCRIPTION:Many algorithms for inferring causality are based
on partial correlation testing. Partial correlatio
ns define hypersurfaces in the parameter space of
a directed Gaussian graphical model. The volumes o
btained by bounding partial correlations play an i
mportant role for the performance of causal infere
nce algorithms. By computing these volumes we show
that the so-called "strong-faithfulness assumptio
n"\, one of the main constraints of many causal in
ference algorithms\, is in fact extremely restrict
ive\, implying fundamental limitations for these a
lgorithms. We then propose an alternative method t
hat involves finding the permutation of the variab
les that yields the sparsest DAG. In the Gaussian
setting\, our sparsest permutation (SP) algorithm
boils down to determining the permutation with spa
rsest Cholesky decomposition of the inverse covari
ance matrix. We prove that the constraints require
d for our SP algorithm are strictly weaker than st
rong-faithfulness and are necessary for any causal
inference algorithm based on conditional independ
ence testing.
LOCATION:MR12\, Centre for Mathematical Sciences\, Wilberf
orce Road\, Cambridge
CONTACT:
END:VEVENT
END:VCALENDAR