BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Dormant independence - Dr Ricardo Silva\, Dept of Statistics\, UCL
DTSTART:20090302T163000Z
DTEND:20090302T173000Z
UID:TALK16651@talks.cam.ac.uk
CONTACT:Dr Clive Bowsher
DESCRIPTION:The construction of causal graphs from non-experimental data r
 ests on a set of constraints that the graph structure imposes on all proba
 bility distributions compatible with the graph. These constraints are of t
 wo types: conditional independencies and algebraic constraints\, first not
 ed by Verma. While conditional independencies are well studied and frequen
 tly used in causal induction algorithms\, Verma constraints are still poor
 ly understood\, and rarely applied.\nThis paper examines a special subset 
 of Verma constraints which are easy to understand\, easy to identify and e
 asy to apply\; they arise from dormant independencies\, namely\, condition
 al independencies that hold in interventional distributions.\nA complete a
 lgorithm is given for determining if a dormant independence between two se
 ts of variables is entailed by the causal graph\, such that this independe
 nce is identifiable\, in other words if it resides in an interventional di
 stribution that can be predicted without resorting to interventions. The u
 sefulness of dormant independencies is shown in model testing and inductio
 n by giving an algorithm that uses constraints entailed by dormant indepen
 dencies to prune extraneous edges from a given causal graph.\n\n'Dormant i
 ndependence'\, I. Shpitset and J. Pearl\, UCLA Cognitive Systems Laborator
 y\, Technical Report (R-340)\, April 2008. In Proceedings of the Twenty-Th
 ird Conference on Artificial Intelligence\, 1081-1087\, 2008 http://bayes.
 cs.ucla.edu/csl_papers.html\n
LOCATION:Centre for Mathematical Sciences\, MR15
END:VEVENT
END:VCALENDAR