In this video we introduce another graphical formalism for representing probability distributions called factor graphs: bipartite graphs containing variables and the factors that relate them. We show how to convert directed and undirected graphs to factor graphs, and how doing so can sometimes eliminate loops, supporting efficient inference.
