BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Scalable Causal Discovery for Statistically Efficient Causal Infer
 ence - Sara Magliacane (Universiteit van Amsterdam)
DTSTART:20260122T094500Z
DTEND:20260122T103000Z
UID:TALK241819@talks.cam.ac.uk
DESCRIPTION:Causal discovery methods can identify valid adjustment sets fo
 r causal effect estimation for a small set of target variables\, even when
  the underlying causal graph is unknown. Global causal discovery methods f
 ocus on learning the whole causal graph and therefore enable the recovery 
 of optimal adjustment sets\, i.e.\, sets with the lowest asymptotic varian
 ce\, but they quickly become computationally prohibitive as the number of 
 variables grows. Local causal discovery methods offer a more scalable alte
 rnative by focusing on the local neighborhood of the target variables\, bu
 t they are restricted to statistically suboptimal adjustment sets.&nbsp\;\
 nIn this talk\, I will present two recent methods that combine the computa
 tional efficiency of local methods with the statistical optimality of glob
 al causal discovery methods. First\, I will describe the Sequential Non-An
 cestor Pruning (SNAP) framework (https://arxiv.org/abs/2502.07857). SNAP p
 rogressively identifies and prunes definite non-ancestors of the target va
 riables during the causal discovery process. We show that the resulting su
 bgraph is sufficient for identifying the causal relations between the targ
 ets and their efficient adjustment sets. Then\, I will introduce Local Opt
 imal Adjustments Discovery (LOAD) (https://arxiv.org/abs/2502.07857)\, a m
 ethod for identifying optimal adjustment sets from local information. As a
  first step\, LOAD identifies the causal relation between the targets and 
 tests if the causal effect is identifiable by using only local information
 . If it is identifiable\, it then finds the optimal adjustment set by leve
 raging local causal discovery to infer the mediators and their parents. Ot
 herwise\, it returns the locally valid parent adjustment sets based on the
  learned local structure. For both methods\, I will show that on our evalu
 ation they outperform global methods in scalability\, while providing more
  accurate effect estimation than local methods.
LOCATION:Seminar Room 1\, Newton Institute
END:VEVENT
END:VCALENDAR