BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Robust Counterfactual Inference in Markov Decision Processes - hel
 d in Moller 2 - Milad  Kazemi (King's College London)
DTSTART:20251120T160000Z
DTEND:20251120T163000Z
UID:TALK241192@talks.cam.ac.uk
DESCRIPTION:Reinforcement learning (RL) is increasingly being used to supp
 ort human decision-making in real-world systems. Before deploying RL-learn
 t policies\, we must verify their safety\, particularly in safety-critical
  domains like healthcare. Counterfactual inference enables offline policy 
 evaluation by predicting how an observed sequence of states and actions (u
 nder an existing policy) would have evolved under an alternative policy. H
 owever\, existing counterfactual inference approaches for MDPs assume a fi
 xed causal model of the underlying system\, limiting the validity (and use
 fulness) of counterfactual inference. We relax these assumptions by comput
 ing exact bounds for the counterfactual probabilities across all causal mo
 dels\, leading to more reliable counterfactual analysis. Moreover\, we pro
 ve closed-form expressions for these bounds\, making computation highly ef
 ficient and scalable for handling large-scale MDPs.\nBio: Milad Kazemi is 
 a postdoctoral researcher at the Department of Informatics\, King's Colleg
 e London. He completed his Ph.D. in Computer Science at Newcastle Universi
 ty in 2023. His research focuses on the intersection of control theory\, r
 einforcement learning\, formal methods\, and causality\, with an emphasis 
 on counterfactual analysis and policy synthesis in sequential decision-mak
 ing and safety-critical systems
LOCATION:External
END:VEVENT
END:VCALENDAR