Recursive Partitioning for Personalization using Observational Data

Nathan Kallus

Abstract: We study the problem of learning to choose from $m$ discrete treatment options (e.g., news item or medical drug) the one with best causal effect for a particular instance (e.g., user or patient) wh...