BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Causal Discovery in Network Data - Elena Zheleva (University of Il
 linois Chicago)
DTSTART:20260306T103000Z
DTEND:20260306T111500Z
UID:TALK244429@talks.cam.ac.uk
DESCRIPTION:Most existing causal discovery algorithms rely on the assumpti
 ons that data are independent and identically distributed (i.i.d.). Howeve
 r\, many real-world domains\, such as biological and social networks\, vio
 late the i.i.d. assumption and consist of interacting entities whose attri
 butes exhibit complex relational and causal dependencies\, breaking the SU
 TVA assumption and leading to interference. To address these challenges an
 d to facilitate causal reasoning in network settings\, I will present two 
 recent contributions that develop graphical models and algorithms for caus
 al discovery in network data in the presence of cycles and latent variable
 s. The first contribution introduces relational acyclification\, an operat
 ion specifically designed for cyclic relational causal models that enables
  formal analysis of identifiability in cyclic relational structures. Under
  the assumptions of relational acyclification and &sigma\;-faithfulness\, 
 we establish that the Relational Causal Discovery (RCD) algorithm (Maier e
 t al.\, 2013) is sound and complete for models containing cyclic dependenc
 ies. The second contribution presents RelFCI\, a causal discovery algorith
 m that is sound and complete for relational data subject to latent confoun
 ding. In this work\, we further derive soundness and completeness guarante
 es for relational d-separation in the presence of latent variables\, there
 by extending causal discovery theory to a broader class of relational syst
 ems.
LOCATION:Seminar Room 1\, Newton Institute
END:VEVENT
END:VCALENDAR