Causal Inference: Learning Graph Structures

This lecture covers the topic of causal inference, focusing on learning graph structures to perform causal reasoning. The instructor explains the SGS algorithm, which involves learning the skeleton and orientation of a directed acyclic graph (DAG). The lecture discusses assumptions, such as no latent variables, and details the two-phase process of skeleton learning and edge orientation. Various tests are presented, including conditional dependence and independence tests, to eliminate edges and correctly orient the graph. The goal is to infer causal relationships from observational data, with examples and practical applications.