Machine Learning in Production: Developing and Optimizing Data Science Workflows and Applications

11. Bayesian Networks

11.1 Introduction

Graphical models are a rich subject, and several excellent books focus on them. Koller and Friedman’s Probabilistic Graphical Models covers the subject in-depth, and Murphy’s Machine Learning: A Probabilistic Perspective uses them to teach machine learning. Graphical models provide a powerful framework for many modern algorithms for matrix factorization and text analysis.

Graphical models are what they sound like: models for systems that are described by graphs, that is, nodes and edges. Each node is a variable, and each edge represents dependence between variables. These models can be directed or undirected. Along with the graph, there is generally a list of probability distributions on each of the ...